HoML_Spam_clf.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Spam Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Exercise : Build a spam classifier with a fair trade-off between precision and recall**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From A. Guéron, \"Hands-On Machine Learning with Scikit-Learn & TensorFlow\", O'Reilly, 2017.\n",
    "<br>\n",
    "Chapter 3 - Classification. Notebook of the author available \n",
    "[here](https://github.com/ageron/handson-ml/blob/master/03_classification.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *Key skills* : Text analysis, bag of words, regular expression, ntlk, binary classification**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Set up**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Importing librairies**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[nltk_data] Downloading package punkt to\n",
      "[nltk_data]     C:\\Users\\ghare\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package punkt is already up-to-date!\n",
      "[nltk_data] Downloading package wordnet to\n",
      "[nltk_data]     C:\\Users\\ghare\\AppData\\Roaming\\nltk_data...\n",
      "[nltk_data]   Package wordnet is already up-to-date!\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "import os\n",
    "import tarfile\n",
    "from six.moves import urllib\n",
    "import re\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from IPython.display import display #to have nice dataframe in Jupyter notebook\n",
    "\n",
    "import sklearn\n",
    "from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer\n",
    "from sklearn.linear_model import SGDClassifier, LogisticRegression\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.naive_bayes import BernoulliNB, MultinomialNB\n",
    "from sklearn.svm import SVC, LinearSVC\n",
    "from sklearn.model_selection import cross_val_predict, cross_val_score, GridSearchCV, train_test_split\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "from sklearn.metrics import precision_recall_curve, recall_score, precision_score, accuracy_score\n",
    "\n",
    "import nltk\n",
    "nltk.download('punkt')\n",
    "nltk.download('wordnet')\n",
    "from nltk import word_tokenize          \n",
    "from nltk.stem import WordNetLemmatizer \n",
    "from nltk.stem.snowball import EnglishStemmer\n",
    "\n",
    "import warnings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Python interpreter :  3.7.3 (default, Apr 24 2019, 15:29:51) [MSC v.1915 64 bit (AMD64)]\n",
      "Numpy version:  1.16.3\n",
      "Pandas version:  0.24.2\n",
      "Seaborn version:  0.9.0\n",
      "Matplotlib version:  3.0.3\n",
      "Scikit-learn version:  0.20.3\n",
      "NTLK version:  3.4.1\n"
     ]
    }
   ],
   "source": [
    "print(\"Python interpreter : \", sys.version)\n",
    "print(\"Numpy version: \", np.__version__)\n",
    "print(\"Pandas version: \", pd.__version__)\n",
    "print(\"Seaborn version: \", sns.__version__)\n",
    "print(\"Matplotlib version: \", matplotlib.__version__)\n",
    "print(\"Scikit-learn version: \", sklearn.__version__)\n",
    "print(\"NTLK version: \", nltk.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.style.use('ggplot')\n",
    "pd.set_option('display.width', 100)\n",
    "pd.set_option('display.max_columns', 30)\n",
    "pd.set_option('display.max_colwidth', 500)\n",
    "warnings.simplefilter(\"ignore\", category=FutureWarning)\n",
    "MODEL_SAVING_PATH = \"./Models/\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Downloading the data**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "DOWNLOAD_ROOT = 'https://spamassassin.apache.org/old/publiccorpus/'\n",
    "FILE_NAMES = ['20030228_easy_ham.tar.bz2', \n",
    "              '20030228_easy_ham_2.tar.bz2', \n",
    "              '20030228_hard_ham.tar.bz2',\n",
    "              '20030228_spam.tar.bz2',\n",
    "              '20050311_spam_2.tar.bz2']\n",
    "SPAM_PATH = os.path.join('datasets', 'spam')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def fetch_spam_data(dl_root=DOWNLOAD_ROOT, file_path=SPAM_PATH):\n",
    "    if not os.path.isdir(file_path):\n",
    "        os.makedirs(file_path)\n",
    "    for file_name in FILE_NAMES:\n",
    "        dl_url = dl_root + file_name\n",
    "        bz2_path = os.path.join(file_path, file_name)\n",
    "        urllib.request.urlretrieve(dl_url, bz2_path)\n",
    "        print(\"{} has been successfully donwloaded on {}\".format(file_name, dl_root))\n",
    "        with tarfile.open(bz2_path, mode='r:bz2') as bz2_file:\n",
    "            bz2_file.extractall(path=file_path)\n",
    "        print(\"It has been successfully unzipped in {}\".format(file_path))\n",
    "        print('~-'*30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "20030228_easy_ham.tar.bz2 has been successfully donwloaded on https://spamassassin.apache.org/old/publiccorpus/\n",
      "It has been successfully unzipped in datasets\\spam\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "20030228_easy_ham_2.tar.bz2 has been successfully donwloaded on https://spamassassin.apache.org/old/publiccorpus/\n",
      "It has been successfully unzipped in datasets\\spam\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "20030228_hard_ham.tar.bz2 has been successfully donwloaded on https://spamassassin.apache.org/old/publiccorpus/\n",
      "It has been successfully unzipped in datasets\\spam\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "20030228_spam.tar.bz2 has been successfully donwloaded on https://spamassassin.apache.org/old/publiccorpus/\n",
      "It has been successfully unzipped in datasets\\spam\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "20050311_spam_2.tar.bz2 has been successfully donwloaded on https://spamassassin.apache.org/old/publiccorpus/\n",
      "It has been successfully unzipped in datasets\\spam\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n"
     ]
    }
   ],
   "source": [
    "fetch_spam_data()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "More info on the data could be found in the readme note [here](https://spamassassin.apache.org/old/publiccorpus/readme.html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['easy_ham', 'easy_ham_2', 'hard_ham', 'spam', 'spam_2'])\n"
     ]
    }
   ],
   "source": [
    "email_type = ['easy_ham', 'easy_ham_2', 'hard_ham', 'spam', 'spam_2']\n",
    "emails_d = {}\n",
    "for type_ in email_type:\n",
    "    email_names = os.listdir(os.path.join(SPAM_PATH, type_))\n",
    "    email_names.remove('cmds')\n",
    "    emails_d[type_] = email_names\n",
    "print(emails_d.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_email(email_path, email_name, encoding='utf8'):\n",
    "    with open(os.path.join(email_path, email_name), 'rb') as file: #rb, issues with specific char\n",
    "        email = file.read().decode(encoding, errors='replace') #back to utf-8 and managing errors\n",
    "    return email\n",
    "     \n",
    "some_email = read_email(email_path=os.path.join(SPAM_PATH, 'easy_ham'), \n",
    "                        email_name='00002.9c4069e25e1ef370c078db7ee85ff9ac')\n",
    "some_email_encoded = some_email.encode()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "b'From Steve_Burt@cursor-system.com  Thu Aug 22 12:46:39 2002\\nReturn-Path: <Steve_Burt@cursor-system.com>\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id BE12E43C34\\n\\tfor <zzzz@localhost>; Thu, 22 Aug 2002 07:46:38 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:46:38 +0100 (IST)\\nReceived: from n20.grp.scd.yahoo.com (n20.grp.scd.yahoo.com\\n    [66.218.66.76]) by dogma.slashnull.org (8.11.6/8.11.6) with SMTP id\\n    g7MBkTZ05087 for <zzzz@spamassassin.taint.org>; Thu, 22 Aug 2002 12:46:29 +0100\\nX-Egroups-Return: sentto-2242572-52726-1030016790-zzzz=spamassassin.taint.org@returns.groups.yahoo.com\\nReceived: from [66.218.67.196] by n20.grp.scd.yahoo.com with NNFMP;\\n    22 Aug 2002 11:46:30 -0000\\nX-Sender: steve.burt@cursor-system.com\\nX-Apparently-To: zzzzteana@yahoogroups.com\\nReceived: (EGP: mail-8_1_0_1); 22 Aug 2002 11:46:29 -0000\\nReceived: (qmail 11764 invoked from network); 22 Aug 2002 11:46:29 -0000\\nReceived: from unknown (66.218.66.217) by m3.grp.scd.yahoo.com with QMQP;\\n    22 Aug 2002 11:46:29 -0000\\nReceived: from unknown (HELO mailgateway.cursor-system.com) (62.189.7.27)\\n    by mta2.grp.scd.yahoo.com with SMTP; 22 Aug 2002 11:46:29 -0000\\nReceived: from exchange1.cps.local (unverified) by\\n    mailgateway.cursor-system.com (Content Technologies SMTPRS 4.2.10) with\\n    ESMTP id <T5cde81f695ac1d100407d@mailgateway.cursor-system.com> for\\n    <forteana@yahoogroups.com>; Thu, 22 Aug 2002 13:14:10 +0100\\nReceived: by exchange1.cps.local with Internet Mail Service (5.5.2653.19)\\n    id <PXX6AT23>; Thu, 22 Aug 2002 12:46:27 +0100\\nMessage-Id: <5EC2AD6D2314D14FB64BDA287D25D9EF12B4F6@exchange1.cps.local>\\nTo: \"\\'zzzzteana@yahoogroups.com\\'\" <zzzzteana@yahoogroups.com>\\nX-Mailer: Internet Mail Service (5.5.2653.19)\\nX-Egroups-From: Steve Burt <steve.burt@cursor-system.com>\\nFrom: Steve Burt <Steve_Burt@cursor-system.com>\\nX-Yahoo-Profile: pyruse\\nMIME-Version: 1.0\\nMailing-List: list zzzzteana@yahoogroups.com; contact\\n    forteana-owner@yahoogroups.com\\nDelivered-To: mailing list zzzzteana@yahoogroups.com\\nPrecedence: bulk\\nList-Unsubscribe: <mailto:zzzzteana-unsubscribe@yahoogroups.com>\\nDate: Thu, 22 Aug 2002 12:46:18 +0100\\nSubject: [zzzzteana] RE: Alexander\\nReply-To: zzzzteana@yahoogroups.com\\nContent-Type: text/plain; charset=US-ASCII\\nContent-Transfer-Encoding: 7bit\\n\\nMartin A posted:\\nTassos Papadopoulos, the Greek sculptor behind the plan, judged that the\\n limestone of Mount Kerdylio, 70 miles east of Salonika and not far from the\\n Mount Athos monastic community, was ideal for the patriotic sculpture. \\n \\n As well as Alexander\\'s granite features, 240 ft high and 170 ft wide, a\\n museum, a restored amphitheatre and car park for admiring crowds are\\nplanned\\n---------------------\\nSo is this mountain limestone or granite?\\nIf it\\'s limestone, it\\'ll weather pretty fast.\\n\\n------------------------ Yahoo! Groups Sponsor ---------------------~-->\\n4 DVDs Free +s&p Join Now\\nhttp://us.click.yahoo.com/pt6YBB/NXiEAA/mG3HAA/7gSolB/TM\\n---------------------------------------------------------------------~->\\n\\nTo unsubscribe from this group, send an email to:\\nforteana-unsubscribe@egroups.com\\n\\n \\n\\nYour use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ \\n\\n\\n\\n'\n"
     ]
    }
   ],
   "source": [
    "print(some_email_encoded)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "From Steve_Burt@cursor-system.com  Thu Aug 22 12:46:39 2002\n",
      "Return-Path: <Steve_Burt@cursor-system.com>\n",
      "Delivered-To: zzzz@localhost.netnoteinc.com\n",
      "Received: from localhost (localhost [127.0.0.1])\n",
      "\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id BE12E43C34\n",
      "\tfor <zzzz@localhost>; Thu, 22 Aug 2002 07:46:38 -0400 (EDT)\n",
      "Received: from phobos [127.0.0.1]\n",
      "\tby localhost with IMAP (fetchmail-5.9.0)\n",
      "\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:46:38 +0100 (IST)\n",
      "Received: from n20.grp.scd.yahoo.com (n20.grp.scd.yahoo.com\n",
      "    [66.218.66.76]) by dogma.slashnull.org (8.11.6/8.11.6) with SMTP id\n",
      "    g7MBkTZ05087 for <zzzz@spamassassin.taint.org>; Thu, 22 Aug 2002 12:46:29 +0100\n",
      "X-Egroups-Return: sentto-2242572-52726-1030016790-zzzz=spamassassin.taint.org@returns.groups.yahoo.com\n",
      "Received: from [66.218.67.196] by n20.grp.scd.yahoo.com with NNFMP;\n",
      "    22 Aug 2002 11:46:30 -0000\n",
      "X-Sender: steve.burt@cursor-system.com\n",
      "X-Apparently-To: zzzzteana@yahoogroups.com\n",
      "Received: (EGP: mail-8_1_0_1); 22 Aug 2002 11:46:29 -0000\n",
      "Received: (qmail 11764 invoked from network); 22 Aug 2002 11:46:29 -0000\n",
      "Received: from unknown (66.218.66.217) by m3.grp.scd.yahoo.com with QMQP;\n",
      "    22 Aug 2002 11:46:29 -0000\n",
      "Received: from unknown (HELO mailgateway.cursor-system.com) (62.189.7.27)\n",
      "    by mta2.grp.scd.yahoo.com with SMTP; 22 Aug 2002 11:46:29 -0000\n",
      "Received: from exchange1.cps.local (unverified) by\n",
      "    mailgateway.cursor-system.com (Content Technologies SMTPRS 4.2.10) with\n",
      "    ESMTP id <T5cde81f695ac1d100407d@mailgateway.cursor-system.com> for\n",
      "    <forteana@yahoogroups.com>; Thu, 22 Aug 2002 13:14:10 +0100\n",
      "Received: by exchange1.cps.local with Internet Mail Service (5.5.2653.19)\n",
      "    id <PXX6AT23>; Thu, 22 Aug 2002 12:46:27 +0100\n",
      "Message-Id: <5EC2AD6D2314D14FB64BDA287D25D9EF12B4F6@exchange1.cps.local>\n",
      "To: \"'zzzzteana@yahoogroups.com'\" <zzzzteana@yahoogroups.com>\n",
      "X-Mailer: Internet Mail Service (5.5.2653.19)\n",
      "X-Egroups-From: Steve Burt <steve.burt@cursor-system.com>\n",
      "From: Steve Burt <Steve_Burt@cursor-system.com>\n",
      "X-Yahoo-Profile: pyruse\n",
      "MIME-Version: 1.0\n",
      "Mailing-List: list zzzzteana@yahoogroups.com; contact\n",
      "    forteana-owner@yahoogroups.com\n",
      "Delivered-To: mailing list zzzzteana@yahoogroups.com\n",
      "Precedence: bulk\n",
      "List-Unsubscribe: <mailto:zzzzteana-unsubscribe@yahoogroups.com>\n",
      "Date: Thu, 22 Aug 2002 12:46:18 +0100\n",
      "Subject: [zzzzteana] RE: Alexander\n",
      "Reply-To: zzzzteana@yahoogroups.com\n",
      "Content-Type: text/plain; charset=US-ASCII\n",
      "Content-Transfer-Encoding: 7bit\n",
      "\n",
      "Martin A posted:\n",
      "Tassos Papadopoulos, the Greek sculptor behind the plan, judged that the\n",
      " limestone of Mount Kerdylio, 70 miles east of Salonika and not far from the\n",
      " Mount Athos monastic community, was ideal for the patriotic sculpture. \n",
      " \n",
      " As well as Alexander's granite features, 240 ft high and 170 ft wide, a\n",
      " museum, a restored amphitheatre and car park for admiring crowds are\n",
      "planned\n",
      "---------------------\n",
      "So is this mountain limestone or granite?\n",
      "If it's limestone, it'll weather pretty fast.\n",
      "\n",
      "------------------------ Yahoo! Groups Sponsor ---------------------~-->\n",
      "4 DVDs Free +s&p Join Now\n",
      "http://us.click.yahoo.com/pt6YBB/NXiEAA/mG3HAA/7gSolB/TM\n",
      "---------------------------------------------------------------------~->\n",
      "\n",
      "To unsubscribe from this group, send an email to:\n",
      "forteana-unsubscribe@egroups.com\n",
      "\n",
      " \n",
      "\n",
      "Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ \n",
      "\n",
      "\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(some_email)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Reading and preparing the data**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 6046 entries, 0 to 6045\n",
      "Data columns (total 3 columns):\n",
      "email    6046 non-null object\n",
      "type     6046 non-null object\n",
      "spam     6046 non-null int64\n",
      "dtypes: int64(1), object(2)\n",
      "memory usage: 141.8+ KB\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "df = pd.DataFrame(columns=['email', 'type'])\n",
    "for k, v in emails_d.items():\n",
    "    email_path = os.path.join(SPAM_PATH, k)\n",
    "    emails_l = [read_email(email_path, name) for name in v]\n",
    "    sub_df = pd.DataFrame({'email': emails_l})\n",
    "    sub_df['type'] = k\n",
    "    df = df.append(sub_df)\n",
    "df['spam'] = df['type'].map(lambda x: 1 if x in('spam', 'spam_2') else 0) \n",
    "df = df.reset_index(drop=True)\n",
    "print(df.info())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Short EDA**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>email</th>\n",
       "      <th>type</th>\n",
       "      <th>spam</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>From exmh-workers-admin@redhat.com  Thu Aug 22 12:36:23 2002\\nReturn-Path: &lt;exmh-workers-admin@spamassassin.taint.org&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id D03E543C36\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 07:36:16 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:36:16 +0100 (IS...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>From Steve_Burt@cursor-system.com  Thu Aug 22 12:46:39 2002\\nReturn-Path: &lt;Steve_Burt@cursor-system.com&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id BE12E43C34\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 07:46:38 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:46:38 +0100 (IST)\\nReceived: ...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>From timc@2ubh.com  Thu Aug 22 13:52:59 2002\\nReturn-Path: &lt;timc@2ubh.com&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id 0314547C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 08:52:58 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 13:52:59 +0100 (IST)\\nReceived: from n16.grp.scd.yahoo.com (n1...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>From irregulars-admin@tb.tf  Thu Aug 22 14:23:39 2002\\nReturn-Path: &lt;irregulars-admin@tb.tf&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id 9DAE147C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 09:23:38 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 14:23:38 +0100 (IST)\\nReceived: from web.tb....</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>From Stewart.Smith@ee.ed.ac.uk  Thu Aug 22 14:44:26 2002\\nReturn-Path: &lt;Stewart.Smith@ee.ed.ac.uk&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id EC69D47C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 09:44:25 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 14:44:25 +0100 (IST)\\nReceived: from n...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>email</th>\n",
       "      <th>type</th>\n",
       "      <th>spam</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>From exmh-workers-admin@redhat.com  Thu Aug 22 12:36:23 2002\\nReturn-Path: &lt;exmh-workers-admin@spamassassin.taint.org&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id D03E543C36\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 07:36:16 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:36:16 +0100 (IS...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>From Steve_Burt@cursor-system.com  Thu Aug 22 12:46:39 2002\\nReturn-Path: &lt;Steve_Burt@cursor-system.com&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id BE12E43C34\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 07:46:38 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 12:46:38 +0100 (IST)\\nReceived: ...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>From timc@2ubh.com  Thu Aug 22 13:52:59 2002\\nReturn-Path: &lt;timc@2ubh.com&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id 0314547C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 08:52:58 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 13:52:59 +0100 (IST)\\nReceived: from n16.grp.scd.yahoo.com (n1...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>From irregulars-admin@tb.tf  Thu Aug 22 14:23:39 2002\\nReturn-Path: &lt;irregulars-admin@tb.tf&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id 9DAE147C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 09:23:38 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 14:23:38 +0100 (IST)\\nReceived: from web.tb....</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>From Stewart.Smith@ee.ed.ac.uk  Thu Aug 22 14:44:26 2002\\nReturn-Path: &lt;Stewart.Smith@ee.ed.ac.uk&gt;\\nDelivered-To: zzzz@localhost.netnoteinc.com\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id EC69D47C66\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 09:44:25 -0400 (EDT)\\nReceived: from phobos [127.0.0.1]\\n\\tby localhost with IMAP (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 14:44:25 +0100 (IST)\\nReceived: from n...</td>\n",
       "      <td>easy_ham</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>email</th>\n",
       "      <th>type</th>\n",
       "      <th>spam</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4150</th>\n",
       "      <td>From 12a1mailbot1@web.de  Thu Aug 22 13:17:22 2002\\nReturn-Path: &lt;12a1mailbot1@web.de&gt;\\nDelivered-To: zzzz@localhost.spamassassin.taint.org\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.spamassassin.taint.org (Postfix) with ESMTP id 136B943C32\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 08:17:21 -0400 (EDT)\\nReceived: from mail.webnote.net [193.120.211.219]\\n\\tby localhost with POP3 (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 13:17:21 +0100 (IS...</td>\n",
       "      <td>spam</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>email</th>\n",
       "      <th>type</th>\n",
       "      <th>spam</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4150</th>\n",
       "      <td>From 12a1mailbot1@web.de  Thu Aug 22 13:17:22 2002\\nReturn-Path: &lt;12a1mailbot1@web.de&gt;\\nDelivered-To: zzzz@localhost.spamassassin.taint.org\\nReceived: from localhost (localhost [127.0.0.1])\\n\\tby phobos.labs.spamassassin.taint.org (Postfix) with ESMTP id 136B943C32\\n\\tfor &lt;zzzz@localhost&gt;; Thu, 22 Aug 2002 08:17:21 -0400 (EDT)\\nReceived: from mail.webnote.net [193.120.211.219]\\n\\tby localhost with POP3 (fetchmail-5.9.0)\\n\\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 13:17:21 +0100 (IS...</td>\n",
       "      <td>spam</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "display(df.head())\n",
    "display(df.loc[df['type'] == 'spam', :].head(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0    0.686404\n",
      "1    0.313596\n",
      "Name: spam, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "print(df['spam'].value_counts(normalize=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax_arr = plt.subplots(1, 2, figsize=(12, 6))\n",
    "sns.countplot(df['type'], order=email_type, ax=ax_arr[0])\n",
    "sns.countplot(df['spam'], palette=['palegreen', 'darkorange'], ax=ax_arr[1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Before going further, let's create and keep aside the test dataset !"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df['email'], df['spam'], stratify=df['spam'], \n",
    "                                                    test_size=0.2, shuffle=True, random_state=42)                                                       \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*Reading back what I have done, here it would be a better call to stratify on df['type'] to ensure a balanced distribution between easy/hard spam/ham*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_exp = X_train.copy()\n",
    "X_spam = X_exp[y_train == 1].reset_index(drop=True)\n",
    "for i in range(0, 101, 25):\n",
    "    print(X_spam[i])\n",
    "    print('~-'*30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's have a quick look on simple and usefull information that might be contained in header e-mail : sender domain, content types and charset ! re library will allow us to play with regular expressions and build the relevant patterns to isolate these information. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "linux.ie\n",
      "text/plain iso-8859-1\n"
     ]
    }
   ],
   "source": [
    "dom_p, cont_p, char_p = 'From .+@(\\S+)',  'Content-Type:\\s?([\\w/]+)\\;', 'charset=\"?([\\w\\-\\_]+)\"?'  \n",
    "get_domain_sender = lambda x: re.findall(dom_p, x)[0].lower() if len(re.findall(dom_p, x)) != 0 else None\n",
    "get_content_type = lambda x: re.findall(cont_p, x)[0].lower() if len(re.findall(cont_p, x)) != 0 else None\n",
    "get_charset = lambda x: re.findall(char_p, x)[0].lower() if len(re.findall(char_p, x)) != 0 else None\n",
    "print(get_domain_sender(X_spam[100]))\n",
    "print(get_content_type(X_spam[100]), get_charset(X_spam[100]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Functions to retrieve domain, content types and charset look to work fine. Let's dig for the full train_set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dom</th>\n",
       "      <th>cont</th>\n",
       "      <th>char</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5644</th>\n",
       "      <td>None</td>\n",
       "      <td>text/html</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2589</th>\n",
       "      <td>linux.ie</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2245</th>\n",
       "      <td>jmason.org</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2706</th>\n",
       "      <td>linux.ie</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1120</th>\n",
       "      <td>freshrpms.net</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>us-ascii</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5475</th>\n",
       "      <td>None</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4417</th>\n",
       "      <td>lists.sourceforge.net</td>\n",
       "      <td>text/html</td>\n",
       "      <td>iso-8859-9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5336</th>\n",
       "      <td>rocketmail.com</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>windows-1252</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3181</th>\n",
       "      <td>redhat.com</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5671</th>\n",
       "      <td>None</td>\n",
       "      <td>text/html</td>\n",
       "      <td>us-ascii</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>dom</th>\n",
       "      <th>cont</th>\n",
       "      <th>char</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5644</th>\n",
       "      <td>None</td>\n",
       "      <td>text/html</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2589</th>\n",
       "      <td>linux.ie</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2245</th>\n",
       "      <td>jmason.org</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2706</th>\n",
       "      <td>linux.ie</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1120</th>\n",
       "      <td>freshrpms.net</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>us-ascii</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5475</th>\n",
       "      <td>None</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>iso-8859-1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4417</th>\n",
       "      <td>lists.sourceforge.net</td>\n",
       "      <td>text/html</td>\n",
       "      <td>iso-8859-9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5336</th>\n",
       "      <td>rocketmail.com</td>\n",
       "      <td>text/plain</td>\n",
       "      <td>windows-1252</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3181</th>\n",
       "      <td>redhat.com</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5671</th>\n",
       "      <td>None</td>\n",
       "      <td>text/html</td>\n",
       "      <td>us-ascii</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "func_l = [get_domain_sender, get_content_type, get_charset]\n",
    "df_exp = pd.concat([X_exp.map(func) for func in func_l], axis=1)\n",
    "df_exp.columns = ['dom', 'cont', 'char']\n",
    "display(df_exp.head(10))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def value_freq_graph(X0, X1, figsize=(10, 5)):\n",
    "    \"\"\" Returns a graph for the value frequencies of 2 Series Ham/Spam from value_counts method \"\"\"\n",
    "    fig, ax_arr = plt.subplots(1, 2, figsize=figsize)\n",
    "    sns.barplot(x= X0.values, y = X0.index, ax=ax_arr[0])\n",
    "    sns.barplot(x= X1.values, y = X1.index, ax=ax_arr[1])\n",
    "    ax_arr[0].set(xlabel='Frequencies', title='Ham')\n",
    "    ax_arr[1].set(xlabel='Frequencies', title='Spam')\n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Domain**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "X0 = df_exp.loc[y_train == 0, 'dom'].value_counts(normalize=True)\n",
    "X1 = df_exp.loc[y_train == 1, 'dom'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "value_freq_graph(X0[:15], X1[:15])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "yahoo.com                     0.079148\n",
       "aol.com                       0.038052\n",
       "insurancemail.net             0.032725\n",
       "msn.com                       0.030441\n",
       "btamail.net.cn                0.019787\n",
       "insiq.us                      0.018265\n",
       "excite.com                    0.016743\n",
       "eudoramail.com                0.015982\n",
       "flashmail.com                 0.015221\n",
       "smtp1.admanmail.com           0.011416\n",
       "netscape.net                  0.010654\n",
       "mail.com                      0.009893\n",
       "caramail.com                  0.007610\n",
       "yahoo.co.uk                   0.006088\n",
       "mail.ru                       0.005327\n",
       "freemail.nl                   0.005327\n",
       "lycos.com                     0.005327\n",
       "email.com                     0.005327\n",
       "frugaljoe.330w.com            0.004566\n",
       "freeuk.com                    0.003805\n",
       "netzero.net                   0.003805\n",
       "internationalfreecall.com     0.003805\n",
       "bluemail.dk                   0.003805\n",
       "firemail.de                   0.003044\n",
       "permissionpass.com            0.003044\n",
       "aeopublishing.com             0.003044\n",
       "ecis.com                      0.003044\n",
       "framesetup.com                0.003044\n",
       "terra.es                      0.003044\n",
       "usa.net                       0.003044\n",
       "                                ...   \n",
       "earthlink.com                 0.000761\n",
       "marketingontarget.net         0.000761\n",
       "icqmail.com                   0.000761\n",
       "missouri.co.jp                0.000761\n",
       "aa64.toplinequotes.com        0.000761\n",
       "coulam1.freeserve.co.uk       0.000761\n",
       "pac6.westernbarge.com         0.000761\n",
       "tcts1.seed.net.tw             0.000761\n",
       "371.net                       0.000761\n",
       "abanet.it                     0.000761\n",
       "abptrade.com.br               0.000761\n",
       "chinchilla.freeserve.co.uk    0.000761\n",
       "fresnomail.com                0.000761\n",
       "none.com                      0.000761\n",
       "tytcorp.com                   0.000761\n",
       "yahoo.com>                    0.000761\n",
       "hushmail.com                  0.000761\n",
       "engfac.uct.ac.za              0.000761\n",
       "loveable.com                  0.000761\n",
       "aemail4u.com                  0.000761\n",
       "theperfumebox.ie              0.000761\n",
       "spl.at                        0.000761\n",
       "yourservice.com               0.000761\n",
       "home.se                       0.000761\n",
       "netscape.com                  0.000761\n",
       "topsitez.us                   0.000761\n",
       "godisenga.com                 0.000761\n",
       "iafrica.com                   0.000761\n",
       "marysstore.com                0.000761\n",
       "swbell.net                    0.000761\n",
       "Name: dom, Length: 462, dtype: float64"
      ]
     },
     "execution_count": 0,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " And what about hotmail.com in ham ?  0.0006523157208088715\n"
     ]
    }
   ],
   "source": [
    "susp_dom = [x for x in X1.index if x not in X0.index]\n",
    "display(X1[susp_dom])\n",
    "print(\"\\n And what about hotmail.com in ham ? \", X0['hotmail.com'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Content type**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "X0 = df_exp.loc[y_train == 0, 'cont'].value_counts(normalize=True)\n",
    "X1 = df_exp.loc[y_train == 1, 'cont'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAFgCAYAAACmDI9oAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3X9UlWW+///Xhi2BhCghpgJCqIGkMojGEI7NiONULk/nWGp+xyxOp7PQkeOIO1BrljqDgqSTlWfSc7JmVdpgqZOtJh06Ocaokz9CEyl/FvIh3PwoMHC7hb2/f7ja694DKKK4AZ+PtVqL+97Xfd3v+zYvX1zc94XJ6XQ6BQAAAECS5OXpAgAAAIDOhIAMAAAAGBCQAQAAAAMCMgAAAGBAQAYAAAAMCMgAAACAAQEZAAAAMCAgAx3siSeeUEpKSoufmUwmvfnmmze5IgDo/i5cuKDnnntOQ4YMkZ+fn+644w6NHj1aL774oqdLQxdg9nQBAAAAN1paWpo+/vhjrVmzRiNHjlRdXZ0+++wzlZaWero0dAHMIAOdxJo1axQXF6fbb79dd955p6ZPn65vvvnG9fmuXbtkMpn0wQcf6Mc//rH8/Pw0atQoFRcXq7i4WMnJyerZs6fGjBmjY8eOefBKAMDztm3bJovFoocffliRkZEaOXKknnjiCf3mN79xtfnhJ3yrV6/WwIED1bNnT02ZMkVVVVWuNocOHdIDDzygkJAQ3X777Ro9erQ+/PBDt3NFREToueeeU1pamgIDAxUSEqKXX35ZFy9e1Ny5c9WnTx8NHDhQL7/88k27flwfAjLQiTz//PP6/PPPtXXrVpWWlmr69OnN2ixevFjZ2dk6ePCgfHx89NhjjyktLU1Lly517XvyySc9UD0AdB79+/fXhx9+qJqamiu2+/TTT7Vr1y59+OGH+uCDD3TkyBGlpqa6Pq+rq9P06dO1a9cuHTp0SBMnTtTkyZN1/Phxt35eeuklDRkyRAcPHlR6errS09P1r//6r4qMjNT+/fv1q1/9Sunp6UxgdBVOAB1q1qxZTm9vb6e/v3+z/yQ533jjjRaPO3TokFOSs6yszOl0Op0ff/yxU5Jz69atrjb5+flOSc533nnHtW/Lli1OSc7z58937IUBQCdWWFjoDA8Pd3p5eTmHDx/u/I//+A/ntm3bnA6Hw9Vm1qxZTn9/f+d3333n2rdjxw6nJOfx48db7XvEiBHO3/3ud67tQYMGOf/lX/7Ftd3U1OQMCAhwTpo0yW1f7969nS+99NKNukR0IGaQgZvg3nvvVVFRUbP/jHbt2qWJEycqLCxMAQEBSk5OliR9/fXXbu1Gjhzp+vrOO++UJI0YMaLZPqvV2iHXAgBdwX333adTp07pk08+0axZs3Tu3DlNmTJFkydPltPpdLUbNmyYAgMD3Y6TpJKSEklSZWWlZs+erejoaPXu3Vu33367iouLrzg2e3l5qW/fvm5js5eXl0JCQhibuwhe0gNuAj8/Pw0ePLjVz0tLS/Xggw9q5syZ+s1vfqPg4GCVlZUpJSVFdrvdrW2PHj1cX5tMplb3ORyOG3kJANDlmM1mJSUlKSkpSRkZGXrzzTc1c+ZM7d69W+PGjWtTH0888YRKS0u1cuVKRUZGys/PT9OnT7/i2CxdHotb2sfY3DUQkIFOYP/+/bpw4YJeeOEF+fn5SZIOHjzo4aoAoHuJiYmR5P4TtpKSEtXV1alXr16SpD179ri13b17t1auXKnJkydLkurr63X69Gndc889N7N03GQ8YgF0AkOGDJHJZNKqVat05swZbdu2TcuWLfN0WQDQZY0bN06vvPKKDhw4oK+//lofffSRZs+erd69e+unP/2pq53JZNLjjz+uo0ePavfu3ZozZ44eeughDRkyRJJ0991366233tLnn3+uoqIiPfbYY2pqavLUZeEmISADncCIESP00ksvad26dRo2bJief/55vfDCC54uCwC6rAceeEBvvfWWHnzwQd1999168sknNWTIEP39739XcHCwq92YMWOUnJysCRMmaOLEiYqNjdVrr73m+vy1116Tw+HQmDFj9PDDD+sXv/iFRo8e7YlLwk1kchqfVAcAALhFPPHEEyorK1NBQYGnS0EnwwwyAAAAYEBABgAAAAx4xAIAAAAwYAYZAAAAMCAgAwAAAAb8opBurLy83NMltFtwcLCqqqo8XUa7Ub9nddb6BwwY4OkS4AGMxZ7FNXheZ6q/reMwM8gAAACAAQEZAAAAMCAgAwAAAAY8g9yNfWN5ytMltNs3ni7gOlG/Z93M+vvn/e9NPBsA4GYgIAMA0IGYrPAsrsHzOrr+jpio4BELAAAAwICADAAAABgQkAEAAAADAjIAAABgQEAGAAAADAjIAAAAgAEBGQAAADAgIAMAAAAGBGQAAADAgIAMAAAAGBCQAQAAAINbPiDX19drx44d7TrWarWqsLCw2f7MzEw1Nja2etzatWu1b9++K/b9pz/9SUeOHGlXXQAAAGg/AnJ9vXbu3NmuYysrK5sFZKvVqqCgIJnN5uuqa9q0aRoxYsR19QEAAIBrd30prhvYuHGjKioqZLFYNGLECAUGBmrv3r26dOmSxowZo6lTp+rkyZN65ZVXtHz5cjkcDi1atEjz5s3Txo0bVVZWJovFonHjxmnSpEkqKipSXFycJGnmzJmaMGGCiouL5e/vr3nz5qlXr15u53/nnXd08OBB2e12DR06VE8//bRMJpPWrl2rUaNGKTExUXPmzNG4ceN08OBBNTY2av78+Ro4cKAnbhcAAEC3d8vPIM+YMUN33nmn8vLyNGLECH3zzTdavny5Vq5cqdOnT+vYsWMaPHiwEhIS9Pbbb+vNN9/U2LFjFR4erhkzZigmJkZ5eXmaNGmSJLkF5IsXLyoyMlK5ubkaNmyYNm/e3Oz8v/jFL7RixQqtWrVKdrtdBw8ebLHOgIAA5ebm6uc//7m2b9/eYpuCggJlZWUpKyvrBt0dAACAW88tP4NsdPjwYR05ckTPPPOMJMlms6miokLDhg3TI488ooULF6pHjx5KTU1t8fjGxkZVV1erX79+kiSTyaSkpCRJ0tixY/X88883O+bo0aN67733dPHiRX3//fcKCwtTQkJCs3b33nuvJOmuu+7Sp59+2uL5U1JSlJKScu0XDgAAABcC8j95+OGHNWHChGb7v//+e9lsNjU2Nsput8vX17dZm5KSEkVHR7fat8lkctu22+169dVXtWLFCgUHBys/P192u73FY394ptnLy0tNTU3XckkAAAC4Brf8IxZ+fn66cOGCJGnkyJH6+OOPZbPZJEk1NTWqra2VJK1bt07Tpk3T2LFj9dZbbzU7Vrr8eMWPfvQj17bT6XStVlFYWNgsPF+6dEmS1KtXL9lsNv3jH//ooKsEAABAW93yM8gBAQG6++67lZGRobi4OCUnJ2vx4sWSJF9fX82dO1dFRUXy9vZWcnKyHA6Hnn32WR09elTR0dHy9vZ2vaR37NgxTZs2zdX3bbfdprNnzyozM1M9e/bUr3/9a7dz+/v7a/z48crIyFBISIiioqJu6rUDAACgOZPT6XR6uojuoLq6WuvWrdOiRYtc+2bOnKk33njDYzUd/P8e9Ni5gVtF/7z/bXPbAQMGdGAlXV99fb0KCws1ceLEaz7WarXq+PHjSk5OdtufmZmp7OxsPfnkky2Ox59++qkGDBig0NDQazrftYzvjMVAx+qIcfiWf8TiRrnjjjvcwjEA4Np4Yl36/fv3q6ysrF3nBNB93fKPWHQkT84eA0BX05Hr0kvSpk2bdOjQIfn4+MhisejcuXM6cOCAjh07pnfffVcZGRl65ZVXFBERoTNnzqiurk5z5szRtm3bVFpaqqSkJE2fPt2DdwjAzUJABgB0CjNmzNDZs2eVl5enw4cPa9++fVq+fLmcTqdWrlypY8eOadiwYa516e12u9u69Nu3b3dbB76oqEizZs2SdHld+iFDhuixxx7Tm2++qY8++khTpkxRQkKC65cy/cBsNmvp0qX64IMPlJeXp5ycHN1+++2aO3euHnroIQUEBFzxOgoKClRQUCBJysnJ6YA7BaCjEZABAJ3OjV6X3mw2a9SoUZIuryd/5MiRVs/9w1r04eHhCg0NVZ8+fSRJ/fr1U3V19VUDMmvSA10fARkA0CndyHXpvb29XWvRX209+R49eki6vHb9D1//sM069MCtgZf0AACdQkeuS9+WcwLAD5hBBgB0Ch25Ln1rkpKStG7dOv3lL3/R/PnzO/oSAXQRrIPcjbH2JtDxWAe582lpXXpPYiwGOhbrIAMAcBWsSw/gehGQAQAAAAOeQe7GruVHDp1NcHCwqqqqPF1Gu1G/Z3X1+gEAnsUMMgAAAGBAQAYAAAAMCMgAAACAAQEZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAA35RSDf2xB/3trj/9Vk/vsmVAAAAdB3MIAMAAAAGBGQAAADAgIAMAAAAGBCQAQAAAAMCMgAAAGDAKhYAAHSg/nn/6+kS2i04OFhVVVWeLuO6cA2e1xXrZwYZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAAwIyAAAAYEBAllRfX68dO3a061ir1arCwsJm+zMzM9XY2KiZM2e2eNynn36qsrKyaz5fa/0BAADgxiAg63JA3rlzZ7uOraysbBaQrVargoKCZDa3vsz0/v372xWQAQAA0LH4RSGSNm7cqIqKClksFo0YMUKBgYHau3evLl26pDFjxmjq1Kk6efKkXnnlFS1fvlwOh0OLFi3SvHnztHHjRpWVlclisWjcuHGaNGmSioqKFBcX5+p/06ZNOnTokHx8fGSxWHTu3DkdOHBAx44d07vvvquMjAy98sorioiI0JkzZ1RXV6c5c+Zo27ZtKi0tVVJSkqZPn+7BOwQAAHDrICBLmjFjhs6ePau8vDwdPnxY+/bt0/Lly+V0OrVy5UodO3ZMw4YNU0JCgt5++23Z7XaNHTtW4eHhmjFjhrZv366srCxXf0VFRZo1a5Yk6eLFixoyZIgee+wxvfnmm/roo480ZcoUJSQkaNSoUUpMTHQdZzabtXTpUn3wwQfKy8tTTk6Obr/9ds2dO1cPPfSQAgICrngdBQUFKigokCTl5OR0wJ0CAADo/gjI/+Tw4cM6cuSInnnmGUmSzWZTRUWFhg0bpkceeUQLFy5Ujx49lJqa2uLxjY2Nqq6uVr9+/SRdDr2jRo2SJN111106cuRIq+dOSEiQJIWHhys0NFR9+vSRJPXr10/V1dVXDcgpKSlKSUm5tgsGAACAGwJyCx5++GFNmDCh2f7vv/9eNptNjY2Nstvt8vX1bdampKRE0dHRrm1vb2+ZTCZJkpeXl5qamlo9b48ePSRJJpPJ9fUP21c6DgDQeT3xx72eLgFo1euzfuzpEjolXtKT5OfnpwsXLkiSRo4cqY8//lg2m02SVFNTo9raWknSunXrNG3aNI0dO1ZvvfVWs2Oly49X/OhHP7qmcwIAAKDzYAZZUkBAgO6++25lZGQoLi5OycnJWrx4sSTJ19dXc+fOVVFRkby9vZWcnCyHw6Fnn31WR48eVXR0tLy9vV0v6R07dkzTpk276jmTkpK0bt06/eUvf9H8+fM7+hIBAADQRian0+n0dBHdRXV1tdatW6dFixZ5uhRJ0s9XvNvi/q7w45Tg4GBVVVV5uox2o37P6qz1DxgwwNMlwANaG4uBzuBmZILONCa3dRzmEYsb6I477ug04RgAAADtQ0AGAAAADAjIAAAAgAEBGQAAADAgIAMAAAAGBGQAAADAgIAMAAAAGBCQAQAAAAMCMgAAAGBAQAYAAAAMzJ4uAB2nK/xKaQAAgM6GGWQAAADAgIAMAAAAGBCQAQAAAAMCMgAAAGBAQAYAeFR9fb127NjRrmOtVqsKCwub7c/MzFRjY2Orx61du1b79u27Yt9/+tOfdOTIkXbVBaBrIyADADyqvr5eO3fubNexlZWVzQKy1WpVUFCQzObrW6hp2rRpGjFixHX1AaBrYpk3AIBHbdy4URUVFbJYLBoxYoQCAwO1d+9eXbp0SWPGjNHUqVN18uRJvfLKK1q+fLkcDocWLVqkefPmaePGjSorK5PFYtG4ceM0adIkFRUVKS4uTpI0c+ZMTZgwQcXFxfL399e8efPUq1cvt/O/8847OnjwoOx2u4YOHaqnn35aJpNJa9eu1ahRo5SYmKg5c+Zo3LhxOnjwoBobGzV//nwNHDjQE7cLwE3ADDIAwKNmzJihO++8U3l5eRoxYoS++eYbLV++XCtXrtTp06d17NgxDR48WAkJCXr77bf15ptvauzYsQoPD9eMGTMUExOjvLw8TZo0SZLcAvLFixcVGRmp3NxcDRs2TJs3b252/l/84hdasWKFVq1aJbvdroMHD7ZYZ0BAgHJzc/Xzn/9c27dv77gbAsDjmEHuxna8902L+ydO7n+TKwGAtjl8+LCOHDmiZ555RpJks9lUUVGhYcOG6ZFHHtHChQvVo0cPpaamtnh8Y2Ojqqur1a9fP0mSyWRSUlKSJGns2LF6/vnnmx1z9OhRvffee7p48aK+//57hYWFKSEhoVm7e++9V5J011136dNPP231GgoKClRQUCBJysnJuYarB9BZEJABAJ3Kww8/rAkTJjTb//3338tms6mxsVF2u12+vr7N2pSUlCg6OrrVvk0mk9u23W7Xq6++qhUrVig4OFj5+fmy2+0tHvvDM81eXl5qampq9RwpKSlKSUlp9XMAnR+PWAAAPMrPz08XLlyQJI0cOVIff/yxbDabJKmmpka1tbWSpHXr1mnatGkaO3as3nrrrWbHSpcfr/jRj37k2nY6na7VKgoLC5uF50uXLkmSevXqJZvNpn/84x8ddJUAuhJmkAEAHhUQEKC7775bGRkZiouLU3JyshYvXixJ8vX11dy5c1VUVCRvb28lJyfL4XDo2Wef1dGjRxUdHS1vb2/XS3rHjh3TtGnTXH3fdtttOnv2rDIzM9WzZ0/9+te/dju3v7+/xo8fr4yMDIWEhCgqKuqmXjuAzsnkdDqdni4CHeO1V1p+0aQrPIMcHBysqqoqT5fRbtTvWZ21/gEDBni6hG6turpa69at06JFi1z7Zs6cqTfeeMODVUk/X/GuR88PXMnrs37c4efoTGNyW8dhHrEAAHQLd9xxh1s4BoD2IiADALotT88eA+iaCMgAAACAAQEZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAgy4ZkGfOnClJslqtKiwsdO0/deqUNmzYcMPOs2XLlmb71q9fry+++KLF9gcOHNC2bdtu2PlbYrValZGR0aHnAAAAuJV1yYD8g8rKSreAHBUVpdTU1Ovu1+l0yuFwaOvWrc0+O3HihIYOHdricQkJCXr44Yev+/wAAADwHLMnTmq1WrV8+XJFR0frxIkTGjRokO6//35t3rxZtbW1Sk9P16FDh+Tr66vJkydLkjIyMpSZmamQkBBXPxs3blRZWZksFovGjRunyMhIbd++XVlZWcrPz9e5c+dUU1Oj6upqTZ48WSkpKbLZbFq5cqXq6+vV2Nio6dOna/To0bJarVqxYoViY2N1/PhxRUREyG63y2KxKCwsTOnp6SorK1P//v3l5eWlDz74QH/961/l7e2t0NBQzZs3T7t27dKpU6f07//+76qoqNBLL70kh8OhuLg4vf/++3rjjTdUXFyszZs3KyAgQGfPntVdd92luXPnymQy6fTp0/rjH/8om82mXr16afbs2erTp49Onz6tP/zhD/Lx8VF0dLQn/sgAAABuGR4JyJJUUVGh+fPnKzQ0VAsXLlRhYaGWLVumAwcOaMuWLYqIiLhqHzNmzHAFYkkqLi52+7y0tFTZ2dmy2WzKzMxUfHy8AgMDtWDBAvXs2VN1dXVavHixEhISJEnl5eVKS0vTU089JUnau3ev8vLyXP0VFRUpLi5OkvTnP/9ZL7/8snr06KH6+vpmtb3++ut64IEHlJycrJ07d7p9dubMGa1evVp9+vTRc889py+//FKDBw/Whg0b9Mwzz6hXr17as2ePNm3apNmzZ+u///u/lZqaqmHDhl3xt0IVFBSooKBAkpSTk3PV+wcAAIDmPBaQQ0JCFB4eLkkKCwvT8OHDZTKZFB4ersrKyjYF5KtJSEiQj4+PfHx8FBsbq5MnTyo+Pl6bNm1SSUmJTCaTampqVFtbK0kKDg5u9fEJSTp8+LDS0tIkSeHh4XrxxRc1evRojRkzplnb48ePy2KxSJKSk5Pdgu3gwYN1xx13SJIiIiJktVrVs2dPnT17Vr/97W8lSQ6HQ3369FFDQ4Pq6+s1bNgwSdJPfvITFRUVtVhfSkqKUlJSrvU2AQAAwMBjAblHjx6ur00mk2vbZDLJ4XDI29tbTqfT1cZut1/zOUwmU7PtwsJC1dXVKScnR2azWXPmzHH17evr22pfFy9eVH19vYKCgiRJCxcu1LFjx3TgwAG9++67Wr16dZvrMl67l5eXHA6HJCk0NFTZ2dlubevr65tdBwAAADpOp31Jr2/fvjpz5owk6fTp07Jarc3a+Pn56cKFC632sX//ftntdp0/f17FxcWKiopSQ0ODAgMDZTabdfToUVVWVrZ6vNlsVmNjo6TLj2/ExsZKujy7W1VVpXvuuUe//OUv1dDQIJvN5nbskCFD9I9//EOStGfPnqte74ABA1RXV6fjx49LkhobG3X27Fn5+/urZ8+erpUzPvnkk6v2BQAAgPbz2Azy1SQmJmr37t2yWCyKiorSgAEDmrUJDw+Xt7e320t6RoMHD1ZOTo6qqqo0ZcoUBQUFKTk5Wbm5ucrKylJERIQGDhzYag3jx4+XxWJRZGSk/P39lZiYKOlyQH7ppZfU0NAgSXrooYfk7+/vduwTTzyhl156Sdu3b1d8fLx69ux5xes1m83KyMjQa6+9poaGBjU1NenBBx9UWFiYZs+e7XpJb+TIkW26fwAAAGgfk9P4HEM3kp+f77YKxvXKzMxUdna2zOa2fU9x8eJF+fj4yGQy6e9//7v+/ve/65lnnrkhtbTVa68cbHH/xMn9b2od7REcHKyqqipPl9Fu1O9ZnbX+lr7RR/f38xXveroEoFWvz/pxh5+jM43JbR2HO+0McmeTm5t7Te1Pnz6tDRs2yOl0yt/f3/VyHwAAADq3bhuQp06d6tHzx8TEuC0RBwC4Nd2MGbqO0plm/tqLa0B7dNqX9AAAAABPICADAAAABgRkAAAAwICADAAAABgQkAEAAAADAjIAAABgQEAGAAAADAjIAAAAgEG3/UUh6Bq/UhoAAKCzYQYZAAAAMGAGGQCADrTjvW88XcJ1uHLt/KQS3RUzyAAAAIABARkAAAAwICADAAAABgRkAAAAwICADAAAABgQkAEAAAADlnnrxl588UW37fT0dA9VAgAA0HUwgwwAAAAYEJABAAAAAwIyAAAAYEBABgAAAAwIyAAAAIABARkAAAAwICADAAAABgRkAAAAwICADAAAABgQkAEAAAADAjIAAABgQEAGAAAADDo8IM+cOVOSZLVaVVhY6Np/6tQpbdiw4YadZ8uWLc32rV+/Xl988UWrx+Tn5+u9996TJO3atUs1NTU3rJ7W1NfXa8eOHa7tmpoarVq1qsPPC+DWxljsjrEYwJXctBnkyspKt0E5KipKqamp192v0+mUw+HQ1q1bm3124sQJDR06tE397Nq1S99+++01nbupqema2kuXB+WdO3e6toOCgpSRkXHN/QBAezAWX8ZYDOBKzFdrYLVatXz5ckVHR+vEiRMaNGiQ7r//fm3evFm1tbVKT0/XoUOH5Ovrq8mTJ0uSMjIylJmZqZCQEFc/GzduVFlZmSwWi8aNG6fIyEht375dWVlZys/P17lz51RTU6Pq6mpNnjxZKSkpstlsWrlyperr69XY2Kjp06dr9OjRslqtWrFihWJjY3X8+HFFRETIbrfLYrEoLCxM6enpKisrU//+/eXl5aWCggJ99NFHamxsVL9+/TR37lzddtttrtr27dunU6dO6cUXX5SPj4+ys7NVVlamP/7xj7LZbOrVq5dmz56tPn36aMmSJRo6dKi+/PJLJSQkqLS0VH5+fjp9+rS+++47/fKXv1RiYmKrtW/cuFEVFRWyWCwaMWKEJk6cqNzcXK1atUqLFi1SWlqawsLCJElLlizR448/rgEDBmjDhg06e/asmpqa9Oijj2r06NE3+v8FAJ0YYzFjMYCb56oBWZIqKio0f/58hYaGauHChSosLNSyZct04MABbdmyRREREVftY8aMGa5BWJKKi4vdPi8tLVV2drZsNpsyMzMVHx+vwMBALViwQD179lRdXZ0WL16shIQESVJ5ebnS0tL01FNPSZL27t2rvLw8V39FRUWKi4uTJN17771KSUmRJL399tv6v//7Pz3wwAOutomJifrwww81c+ZMRUVFqbGxURs2bNAzzzyjXr16ac+ePdq0aZNmz54tSWpoaNDSpUslSWvXrtV3332nZcuWqby8XLm5uUpMTFSPHj1arH3GjBk6e/asq1ar1eqqIykpSXv37lVYWJi+/fZbffvtt7rrrru0ceNG3XPPPZo9e7bq6+u1aNEiDR8+XL6+vm73sKCgQAUFBZKknJycq/6ZAOhaGIsZiwHcHG0KyCEhIQoPD5ckhYWFafjw4TKZTAoPD1dlZWWbBuWrSUhIkI+Pj3x8fBQbG6uTJ08qPj5emzZtUklJiUwmk2pqalRbWytJCg4OvuKP7A4fPqy0tDRJ0tmzZ/X222+rvr5eNptNI0eOvGIt5eXlOnv2rH77299KkhwOh/r06eP6PCkpya396NGj5eXlpdDQUFd9Tqez1dpbk5SUpN/+9reaOnWq9u7dq8TEREnSkSNHdPDgQW3fvl2SZLfbVVVVpdDQULfjU1JSXP/4AOh+GIsZiwHcHG0KyD169HB9bTKZXNsmk0kOh0Pe3t5yOp2uNna7/ZoLMZlMzbYLCwtVV1ennJwcmc1mzZkzx9X3P3/HbnTx4kXV19crKChI0uWZBYvFooiICO3atavZjElLQkNDlZ3uuAwcAAAgAElEQVSd3eJnxh8JSu7354f7cKXaWxMUFKSAgAB9/fXX2rNnj55++mlXnxkZGRowYMBV6wbQfTEWu2MsBtBRbshLen379tWZM2ckSadPn3b7UdUP/Pz8dOHChVb72L9/v+x2u86fP6/i4mJFRUWpoaFBgYGBMpvNOnr0qCorK1s93mw2q7GxUdLlHxnGxsa6PrPZbOrTp48aGxv1ySeftHi8r6+vq74BAwaorq5Ox48flyQ1Njbq7NmzV7kL7lqr/Wr3ISkpSX/+85/V0NDgmikaOXKk/vKXv7gG/B/uNQAYMRY3x1gMoD3aNIN8NYmJidq9e7csFouioqJa/O46PDxc3t7ebi+GGA0ePFg5OTmqqqrSlClTFBQUpOTkZOXm5iorK0sREREaOHBgqzWMHz9eFotFkZGR8vf3d/1ITJKmTZumRYsWqW/fvgoPD29xULz//vv1P//zP64XQzIyMvTaa6+poaFBTU1NevDBB10vbLRFa7UHBATo7rvvVkZGhuLi4jRx4sRm9/L111/XlClTXPseeeQRvf7661qwYIGky/8I/vD8IAD8gLG4OcZiAO1hchp/Huch+fn5bm9eX6/MzExlZ2fLbL4h+b/L+ueBOz093UOVXLvg4GBVVVV5uox2o37P6qz1d/YfzTMWd4zXXjno6RI6zMTJ/T1dwlV11vHgWnT1a+hM9bd1HO6Wo1Zubq6nSwCAWx5jMYCuqlME5KlTp3q6BAC45TEWA8BlN+036QEAAABdAQEZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAAwIyAAAAYEBABgAAAAwIyAAAAIBBp/hNeugY6enpni4BAACgy2EGGQAAADAgIAMAAAAGBGQAAADAgIAMAAAAGBCQAQAAAANWsQAAoANNnNzf0yW0W3BwsKqqqjxdBnDTMYMMAAAAGBCQuzGvklWeLgEAAKDLISADAAAABgRkAAAAwICADAAAABgQkAEAAAADAjIAAABgQEAGAAAADAjIAAAAgAEBGQAAADAgIAMAAAAGZk8XAABAd/biiy/e1POlp6ff1PMB3REzyAAAAIABARkAAAAw6PIBeebMmZIkq9WqwsJC1/5Tp05pw4YNN+w8W7ZsabZv/fr1+uKLL9rcx7PPPntDalm7dq327dt3Q/oCAACAuy4fkH9QWVnpFpCjoqKUmpp63f06nU45HA5t3bq12WcnTpzQ0KFD29zX7373u+uuBwAAAB3L4y/pWa1WLV++XNHR0Tpx4oQGDRqk+++/X5s3b1Ztba3S09N16NAh+fr6avLkyZKkjIwMZWZmKiQkxNXPxo0bVVZWJovFonHjxikyMlLbt29XVlaW8vPzde7cOdXU1Ki6ulqTJ09WSkqKbDabVq5cqfr6ejU2Nmr69OkaPXq0rFarVqxYodjYWB0/flwRERGy2+2yWCwKCwtTenq6ysrK1L9/f3l5eWnJkiWKiIjQmTNnVFdXpzlz5mjbtm0qLS1VUlKSpk+fLunybPcbb7yhTz/9VDt27NCzzz6r7777TkuWLNHSpUvVq1cvvfXWWzp27JguXbqkiRMnasKECXI6ndqwYYOOHj3qds0AAAC48TwekCWpoqJC8+fPV2hoqBYuXKjCwkItW7ZMBw4c0JYtWxQREXHVPmbMmOEKxJJUXFzs9nlpaamys7Nls9mUmZmp+Ph4BQYGasGCBerZs6fq6uq0ePFiJSQkSJLKy8uVlpamp556SpK0d+9e5eXluforKipSXFyca9tsNmvp0qX64IMPlJeXp5ycHN1+++2aO3euHnroIQUEBLjajhkzRvv27dOOHTtUVFSkRx99VL1791ZBQYF69uypFStW6NKlS3ruuec0cuRInTlzRuXl5Vq1apW+++47zZ8/Xz/96U+b3YOCggIVFBRIknJyctp49wEAAGDUKQJySEiIwsPDJUlhYWEaPny4TCaTwsPDVVlZ2aaAfDUJCQny8fGRj4+PYmNjdfLkScXHx2vTpk0qKSmRyWRSTU2NamtrJUnBwcFXfHzi8OHDSktLc+tfksLDwxUaGqo+ffpIkvr166fq6mq3gCxJqampysjI0JAhQ5ScnOzqs7S01PV8cUNDg7755huVlJTovvvuk5eXl4KCgnTPPfe0WFNKSopSUlLaeYcAAAAgdZKA3KNHD9fXJpPJtW0ymeRwOOTt7S2n0+lqY7fbr/kcJpOp2XZhYaHq6uqUk5Mjs9msOXPmuPr29fVtta+LFy+qvr5eQUFBza7BWP8P201NTc36qKmpkZeXl2pra+VwOOTl5SWn06knn3zSbWZakj777LNm9QMAAKBjdImX9Pr27aszZ85Ikk6fPi2r1dqsjZ+fny5cuNBqH/v375fdbtf58+dVXFysqKgoNTQ0KDAwUGazWUePHlVlZWWrx5vNZjU2Nkq6/PhGbGxsu6+nqalJf/jDH5Senq6BAwfq/ffflyTFxcVp586drvOUl5fLZrMpJiZGe/bskcPh0Lffftvs8REAAADcOJ1iBvlqEhMTtXv3blksFkVFRWnAgAHN2oSHh8vb29vtJT2jwYMHKycnR1VVVZoyZYqCgoKUnJys3NxcZWVlKSIiQgMHDmy1hvHjx8tisSgyMlL+/v5KTExs9/Vs3bpV0dHRiomJUUREhBYuXKj4+Hj97Gc/k9VqVWZmpiSpV69eslgsGjNmjI4ePaqMjAz1799fMTEx7T43AAAArszkND670E3l5+e7rYJxvTIzM5WdnS2zuXN/f1HxUYYcMRmeLqNdgoODVVVV5eky2o36Pauz1t/SN/fo/n54efxmuZG/arqz/l26FlyD53Wm+ts6DnfuhNdJ5ebmeroEAAAAdJBbIiBPnTrV0yUAAACgi+gSL+kBAAAANwsBGQAAADAgIAMAAAAGBGQAAADAgIAMAAAAGBCQAQDXZebMmZIkq9WqwsJC1/5Tp05pw4YNN+w8W7ZsabZv/fr1+uKLL9rcx7PPPntDalm7dq327dt3Q/oC0PkQkAEAN0RlZaVbQI6KilJqaup19+t0OuVwOLR169Zmn504cUJDhw5tc1+/+93vrrseAN3fLbEOMgCgdVarVcuXL1d0dLROnDihQYMG6f7779fmzZtVW1ur9PR0HTp0yO03kmZkZCgzM1MhISGufjZu3KiysjJZLBaNGzdOkZGR2r59u7KyspSfn69z586ppqZG1dXVmjx5slJSUmSz2bRy5UrV19ersbFR06dP1+jRo2W1WrVixQrFxsbq+PHjioiIkN1ul8ViUVhYmNLT01VWVqb+/fvLy8tLS5YsUUREhM6cOaO6ujrNmTNH27ZtU2lpqZKSkjR9+nRJl2e733jjDX366afasWOHnn32WX333XdasmSJli5dql69eumtt97SsWPHdOnSJU2cOFETJkyQ0+nUhg0bdPToUbdrBtA9EZABAKqoqND8+fMVGhqqhQsXqrCwUMuWLdOBAwe0ZcsWRUREXLWPGTNmuAKxJBUXF7t9XlpaquzsbNlsNmVmZio+Pl6BgYFasGCBevbsqbq6Oi1evFgJCQmSpPLycqWlpempp56SJO3du1d5eXmu/oqKihQXF+faNpvNWrp0qT744APl5eUpJydHt99+u+bOnauHHnpIAQEBrrZjxozRvn37tGPHDhUVFenRRx9V7969VVBQoJ49e2rFihW6dOmSnnvuOY0cOVJnzpxReXm5Vq1ape+++07z58/XT3/603bfbwCdGwG5G3PEZHi6BABdREhIiMLDwyVJYWFhGj58uEwmk8LDw1VZWdmmgHw1CQkJ8vHxkY+Pj2JjY3Xy5EnFx8dr06ZNKikpkclkUk1NjWprayVJwcHBV3x84vDhw0pLS3PrX5LCw8MVGhqqPn36SJL69eun6upqt4AsSampqcrIyNCQIUOUnJzs6rO0tNT1fHFDQ4O++eYblZSU6L777pOXl5eCgoJ0zz33tFpXQUGBCgoKJEk5OTnXepsAdAIEZACAevTo4fraZDK5tk0mkxwOh7y9veV0Ol1t7Hb7NZ/DZDI12y4sLFRdXZ1ycnJkNps1Z84cV9++vr6t9nXx4kXV19crKCio2TUY6/9hu6mpqVkfNTU18vLyUm1trRwOh7y8vOR0OvXkk0+6zUxL0meffdas/takpKQoJSWlTW0BdE68pAcAuKq+ffvqzJkzkqTTp0/LarU2a+Pn56cLFy602sf+/ftlt9t1/vx5FRcXKyoqSg0NDQoMDJTZbNbRo0dVWVnZ6vFms1mNjY2SLj++ERsb2+7raWpq0h/+8Aelp6dr4MCBev/99yVJcXFx2rlzp+s85eXlstlsiomJ0Z49e+RwOPTtt982e3wEQPfCDDIA4KoSExO1e/duWSwWRUVFacCAAc3ahIeHy9vb2+0lPaPBgwcrJydHVVVVmjJlioKCgpScnKzc3FxlZWUpIiJCAwcObLWG8ePHy2KxKDIyUv7+/kpMTGz39WzdulXR0dGKiYlRRESEFi5cqPj4eP3sZz+T1WpVZmamJKlXr16yWCwaM2aMjh49qoyMDPXv318xMTHtPjeAzs/kNP7MDN1KeXm5p0tot+DgYFVVVXm6jHajfs/qrPW3FCpvFfn5+W6rYFyvzMxMZWdny2zu/PM8P7y0eLOkp6ffsL4669+la8E1eF5nqr+t43DnH1kAAPgnubm5ni4BQDdGQAYAdLipU6d6ugQAaDNe0gMAAAAMCMgAAACAAQEZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAAwIyAAAAYEBABgAAAAwIyAAAAIABARkAAAAwICADAAAABgRkAAAAwICADAAAABh0qYA8c+ZMSZLValVhYaFr/6lTp7Rhw4Ybdp4tW7Y027d+/Xp98cUXbe5j165devXVV6/Ypri4WF9++eU11zdnzhzV1dVd83EAAAC4ui4VkH9QWVnpFpCjoqKUmpp63f06nU45HA5t3bq12WcnTpzQ0KFD3fY1NTVd1/naG5ABAADQccw382RWq1XLly9XdHS0Tpw4oUGDBun+++/X5s2bVVtbq/T0dB06dEi+vr6aPHmyJCkjI0OZmZkKCQlx9bNx40aVlZXJYrFo3LhxioyM1Pbt25WVlaX8/HydO3dONTU1qq6u1uTJk5WSkiKbzaaVK1eqvr5ejY2Nmj59ukaPHi2r1aoVK1YoNjZWx48fV0REhOx2uywWi8LCwpSenq6ysjL1799fXl5eWrJkiYYOHaovv/xSCQkJGjdunNavX6/q6mpJ0qxZsxQdHe123QcOHNCWLVvU2NiogIAAzZ07V3a7XX/961/l5eWlTz75RKmpqRo4cGCLfZ0/f15r1qxRXV2dBg8eLKfTeZP+xAAA1ys9Pd3TJQC4Rjc1IEtSRUWF5s+fr9DQUC1cuFCFhYVatmyZK0RGRERctY8ZM2a4ArF0eSbWqLS0VNnZ2bLZbMrMzFR8fLwCAwO1YMEC9ezZU3V1dVq8eLESEhIkSeXl5UpLS9NTTz0lSdq7d6/y8vJc/RUVFSkuLs613dDQoKVLl0qS1qxZo0mTJik6OlpVVVXKzs7W73//e7d6oqOjlZ2dLZPJpI8++kjvvfeeHn/8cU2YMMHtm4HW+tq8ebOio6P1yCOP6NChQyooKGjxvhQUFLg+y8nJuep9BAAAQHM3PSCHhIQoPDxckhQWFqbhw4fLZDIpPDxclZWVbQrIV5OQkCAfHx/5+PgoNjZWJ0+eVHx8vDZt2qSSkhKZTCbV1NSotrZWkhQcHNzs8Qmjw4cPKy0tzbWdlJTk+vrzzz9XWVmZa7uhoUEXLlxwO76mpkYvvPCCvv32WzU2NrrNhhu11ldJSYkWLFggSYqPj5e/v3+Lx6ekpCglJaXV6wAAAMDV3fSA3KNHD9fXJpPJtW0ymeRwOOTt7e32CIHdbr/mc5hMpmbbhYWFqqurU05Ojsxms+bMmePq29fXt9W+Ll68qPr6egUFBbn23Xbbba6vnU6nsrOz5ePj02ofGzZs0KRJk5SQkKDi4mJt3ry5xXZX6uufrwkAAAAdo9O9pNe3b1+dOXNGknT69GlZrdZmbfz8/JrN0hrt379fdrtd58+fV3FxsaKiotTQ0KDAwECZzWYdPXpUlZWVrR5vNpvV2Ngo6fLjG7Gxsa22HTFihD788EPX9ldffdWsTUNDgytg/+1vf3O7DpvNdtW+YmJi9Mknn0iSPvvsM9XX17daDwAAAK7PTZ9BvprExETt3r1bFotFUVFRGjBgQLM24eHh8vb2dntJz2jw4MHKyclRVVWVpkyZoqCgICUnJys3N1dZWVmKiIjQwIEDW61h/PjxslgsioyMlL+/vxITE1tt++STT+rVV1/VggUL1NTUpJiYGD399NNubR599FGtXr1aQUFBGjJkiCv0jxo1SqtXr9b+/fuVmpraal+PPvqo1qxZo8zMTMXExCg4OPhabikAAACugcnZzZZEyM/Pd3vx7XplZmYqOztbZnOn+17iqsrLyz1dQrsFBwerqqrK02W0G/V7Vmetv6Vv+NH9MRZ7FtfgeZ2p/raOw10v9d1kubm5ni4BAAAAN1G3C8hTp071dAkAAADowjrdS3oAAACAJxGQAQAAAINu94gFAACdiVfJqmtq74jJ6KBKALQVM8gAAACAAQEZAAAAMCAgAwAAAAYEZAAAAMCAgAwAAAAYEJABAAAAAwIyAAAAYEBABgAAAAwIyAAAAIABARkAAAAwICADAAAABgRkAAAAwICADAAAABgQkAEAAAADAjIAAABgQEAGAAAADAjIAAAAgAEBGQAAADAgIAMAAAAGBGQAQJvNnDlTkmS1WlVYWOjaf+rUKW3YsOGGnWfLli3N9q1fv15ffPFFm/vYtWuXXn311Su2KS4u1pdffnnN9c2ZM0d1dXXXfByAroGADAC4ZpWVlW4BOSoqSqmpqdfdr9PplMPh0NatW5t9duLECQ0dOtRtX1NT03Wdr70BGUD3ZvZ0AQCAm8dqtWr58uWKjo7WiRMnNGjQIN1///3avHmzamtrlZ6erkOHDsnX11eTJ0+WJGVkZCgzM1MhISGufjZu3KiysjJZLBaNGzdOkZGR2r59u7KyspSfn69z586ppqZG1dXVmjx5slJSUmSz2bRy5UrV19ersbFR06dP1+jRo2W1WrVixQrFxsbq+PHjioiIkN1ul8ViUVhYmNLT01VWVqb+/fvLy8tLS5Ys0dChQ/Xll18qISFB48aN0/r161VdXS1JmjVrlqKjo92u+8CBA9qyZYsaGxsVEBCguXPnym63669//au8vLz0ySefKDU1VQMHDmyxr/Pnz2vNmjWqq6vT4MGD5XQ6b9KfGABPICADwC2moqJC8+fPV2hoqBYuXKjCwkItW7bMFSIjIiKu2seMGTNcgVi6PBNrVFpaquzsbNlsNmVmZio+Pl6BgYFasGCBevbsqbq6Oi1evFgJCQmSpPLycqWlpempp56SJO3du1d5eXmu/oqKihQXF+fabmho0NKlSyVJa9as0aRJkxQdHa2qqiplZ2fr97//vVs90dHRys7Olslk0kcffaT33ntPjz/+uCZMmOD2zUBrfW3evFnR0dF65JFHdOjQIRUUFLR6bwoKClyf5+TkXPVeAuh8CMgAcIsJCQlReHi4JCksLEzDhw+XyWRSeHi4Kisr2xSQryYhIUE+Pj7y8fFRbGysTp48qfj4eG3atEklJSUymUyqqalRbW2tJCk4OLjZ4xNGhw8fVlpamms7KSnJ9fXnn3+usrIy13ZDQ4MuXLjgdnxNTY1eeOEFffvtt2psbHSbDTdqra+SkhItWLBAkhQfHy9/f/9Wa01JSVFKSkqrnwPo/AjIAHCL6dGjh+trk8nk2jaZTHI4HPL29nZ7hMBut1/zOUwmU7PtwsJC1dXVKScnR2azWXPmzHH17evr22pfFy9eVH19vYKCglz7brvtNtfXTqdT2dnZ8vHxabWPDRs2aNKkSUpISFBxcbE2b97cYrsr9fXP1wSg++IlPQCAm759++rMmTOSpNOnT8tqtTZr4+fn12yW1mj//v2y2+06f/68iouLFRUVpYaGBgUGBspsNuvo0aOqrKxs9Xiz2azGxkZJlx/fiI2NbbXtiBEj9OGHH7q2v/rqq2ZtGhoaXAH7b3/7m9t12Gy2q/YVExOjTz75RJL02Wefqb6+vtV6AHR9zCADANwkJiZq9+7dslgsioqK0oABA5q1CQ8Pl7e3t9tLekaDBw9WTk6OqqqqNGXKFAUFBSk5OVm5ubnKyspSRESEBg4c2GoN48ePl8ViUWRkpPz9/ZWYmNhq2yeffFKvvvqqFixYoKamJsXExOjpp592a/Poo49q9erVCgoK0pAhQ1yhf9SoUVq9erX279+v1NTUVvt69NFHtWbNGmVmZiomJkbBwcHXcksBdDEmZxd/FXfmzJl64403ZLVadfz4cSUnJ0u6vCbn3/72txuy7JB0eU3Of/u3f3Pbt379ev3kJz9p9rZ0R6mvr1dhYaEmTpzYpvbl5eUdXFHHCQ4OVlVVlafLaDfq96zOWn9LQbM7ys/Pd3vx7XplZmYqOztbZnPXnNOp+Cjjmto7Yq6tfUfqrH+XrgXX4Hmdqf62jsPd5hGL7rImZ2scDofq6+u1c+fODukfADqr3NzcLhuOAXRNHh9xbqU1OVurw+l06s0331RRUZEkacqUKUpKSlJxcbHeeecd9e7dW1999ZUGDRqkiooKWSwWjRgxwvUbrQCgM5k6daqnSwCA6+LxgCzdWmtytlTH8ePH9dVXXykvL091dXVauHChYmJiJEknT57UqlWrFBISIqvVqrNnz7rVYcTamwAAANevUwTkW2lNzpbq+OKLL3TffffJy8tLvXv31rBhw3Tq1Cn5+flp8ODBra7X+c9YexMAAOD6dYqAfCutydlSHVdi7BcAAAAdr0u8pNed1uRsqY6YmBjt3btXDodDdXV1Kikp0eDBg6/5GgEAAHD9OsUM8tV0pzU5W6pjzJgxOn78uCwWiyTpl7/8pXr37q3/9//+n1u/AQEBuvvuu5WRkaG4uDhe0gMAAOgAXX4d5LboLGty3ug6roZ1kD2H+j2rs9Z/q6yDDHesg+xZXIPndab62zoOd4kZ5M4mNzfX0yUAAACgg9wSAbmzrMnZWeoAAABA67rES3oAAADAzUJABgAAAAwIyAAAAIABARkAAAAwICADAAAABgRkAAAAwICADAAAABjcEusgAwDgKZ3pN+MBaBtmkAEAAAADAjIAAABgQEAGAAAADAjIAAAAgAEBGQAAADAwOZ1Op6eLAAAAADoLZpC7qaysLE+XcF2o37OoH7gxuvr/i129folr6Ay6Yv0EZAAAAMCAgAwAAAAYeC9ZsmSJp4tAx7jrrrs8XcJ1oX7Pon7gxujq/y929folrqEz6Gr185IeAAAAYMAjFgAAAIABARkAAAAwMHu6AFyfoqIivfbaa3I4HBo/frwefvhht88vXbqkl19+WadPn1ZAQIDmzZunkJAQD1Xb3NXqf//99/XRRx/J29tbvXr1Ulpamvr27euhapu7Wv0/2Ldvn1avXq0VK1YoKirqJlfZurbUv2fPHm3evFkmk0mDBg3Sf/3Xf3mg0pZdrf6qqiqtXbtW9fX1cjgcmjFjhuLj4z1ULbqrrj4OS1e/hmPHjumPf/yjvv76a82bN0+JiYkeqrR13f3fk507d2rHjh3y8vKSr6+v/vM//1OhoaEeqrZlXf3fRDdOdFlNTU3OX/3qV86KigrnpUuXnAsWLHCePXvWrc2HH37oXLdundPpdDoLCwudq1ev9kSpLWpL/Z9//rnTZrM5nU6nc8eOHV2ufqfT6WxoaHD+5je/cS5atMh58uRJD1TasrbUX15e7rRYLM7z5887nU6n8/9v795jqq7/OI4/OdwEEYnDRSjTjWAVRhTismazmdWqscZGuUxC14XEyD8SM2vBYsupWUQ4WpbpcoWX1ZpbtYU1G1kjGCtsLDQimEfgQHjYYciBc35/tE6Hm3ztl+cCr8df5/LhfF/fw3h/3ufz/X45/f39vog6KSP5q6urXV9++aXL5XK5Ojo6XBs3bvRFVJnBAr0Ou1zG9qGrq8v1+++/uyorK12nTp3yUdKpzYb5xG63u2/X19e7ysvLvR3zkgJ9ThxPp1gEsDNnzrBgwQISExMJCQnh9ttvp76+fsyYH3/8kZUrVwJw22230dzcjMtPrss0kn/JkiWEh4cDkJqaSl9fny+iTspIfoCamhpycnIIDQ31QcqpGclfW1vLvffeS1RUFADz58/3RdRJGckfFBTE4OAgAIODg1x11VW+iCozWKDXYTC2DwkJCSxatIigoCAfpby02TCfREZGum8PDQ353e8i0OfE8dQgB7C+vj7MZrP7vtlsnvAH7zkmODiYyMhIBgYGvJpzKkbyezpx4gSZmZneiGaIkfxtbW1YrVaysrK8HW9aRvKfO3cOi8XCyy+/zPbt22lqavJ2zCkZyZ+Xl8e3335LYWEhr732Ghs2bPB2TJnhAr0Ow+XXYn80G+YTgC+++IJnn32WQ4cOsX79em9GnFagz4njqUEOYJOtQIz/RGlkjK9cTraTJ0/y22+/kZOTc6VjGTZdfqfTyYEDB8jPz/dmLMOMvP9OpxOLxcIrr7zCc889R3V1NXa73VsRL8lI/rq6OlauXEl1dTXbtm2jsrISp9PprYgyCwR6HQb/z2fETJ9P/nbfffdRWVnJ2rVrOXbsmDeiGRboc+J4apADmNlspre3132/t7d3wiFkzzGjo6MMDg66D5f7mpH8AD/99BOffPIJJSUlfnVIZrr8Q0NDdHR0UFZWRlFREa2trezcuZOzZ8/6Iu4ERt7/2NhYsrOzCQkJISEhgeTkZCwWi7ejTspI/hMnTrB8+XIA0tLScDgcfrVyJ4Ev0K8s4MkAAAcpSURBVOswGK/F/mymzyfjTXX6gi8F+pw4nhrkAJaSkoLFYqG7u5uRkRG+++47li5dOmZMVlYW33zzDfDXVaPp6el+szJgJH9bWxvvvvsuJSUlfnX+K0yfPzIykvfee4+qqiqqqqpITU2lpKTEb67YNfL+L1u2jObmZgBsNhsWi4XExERfxJ3ASP64uDh3/s7OThwOB9HR0b6IKzNUoNdhMLYP/m6mzyfAmMWJxsZGkpKSvB3zkgJ9ThxP36QX4BobGzlw4ABOp5O77rqL3NxcampqSElJYenSpQwPD/P222/T1tZGVFQUmzdv9psGB6bP/+qrr/LHH38QExMD/NXwbN261cep/zFdfk+lpaWsW7fOr4rBdPldLhcHDx6kqakJk8lEbm4ud9xxh69ju02Xv7Ozk3feeYehoSEAHnvsMW6++WYfp5aZJtDrMEy/D2fOnGH37t3Y7XZCQ0OJiYlhz549vo49xkyfT/bv38/PP/9McHAwUVFRbNiwgYULF/o69hiBPid6UoMsIiIiIuJBp1iIiIiIiHhQgywiIiIi4kENsoiIiIiIBzXIIiIiIiIe1CCLiIiIiHhQgywyi6xbt46uri5fxxARmdVUi/2f/s2byDhFRUX09/djMv3z+bGiooLY2FgfphIRmV1Ui8WXQnwdQMQfbd26lYyMjCmfHx0dJTg42IuJRERmH9Vi8RU1yCIGdHd3s2nTJgoLCzly5AgJCQmUlZXx66+/cvDgQTo7O4mPj6egoID09HT3z1RVVdHW1kZqairJycnY7XaKi4s5ffo0lZWVVFdXu7dRVFTE008/TUZGBk6nk88++4za2lrsdjtLlizhqaeeIioqyp1l48aN1NTUMDw8zAMPPEBubi4ATqeTTz/9lK+//poLFy6QlJTEli1biIuL4+GHH+att95iwYIFOBwOPvroI06dOsXIyAjZ2dkUFBQQFhaGzWZj7969tLS0EBQUxMKFCyktLR2zkiMi4m2qxarF3qIGWeQy/PLLL7zxxhuYTCb6+vrYsWMHmzZtIjMzk+bmZl5//XXefPNNoqOjqaioIC0tjZdeeonW1lZ27Ngx4as2p/L5559TX19PaWkp0dHR7N+/n3379rF582b3mJaWFioqKjh37hwvvvgiy5Yt45prruH48ePU1dWxbds2kpKSaG9vJzw8fMI2Dh06RFdXF7t27SI4OJiKigqOHj3Ko48+yvHjx4mNjWXfvn0AtLa2EhQU9N+8iSIi/yfVYrnS9BFEZBK7du2ioKCAgoICdu7c6X48Ly+POXPmEBYWxsmTJ7nlllu49dZbMZlMZGRkkJKSQmNjI1arlbNnz/LII48QGhrKjTfeSFZWluHtf/XVV6xZswaz2UxoaCh5eXn88MMPjI6OjskSFhbG4sWLWbRoEe3t7QDU1tayZs0akpOTCQoKYvHixcybN2/M67tcLmpra3n88ceJiooiIiKC3Nxc6urqAAgODqa/vx+r1UpISAg33HCDirKIeJ1qsWqxr2gFWWQSW7ZsGXPeW3d3NwBms9n9mNVq5fvvv6ehocH92OjoKOnp6fT19TF37lzmzJnjfi4+Ph6r1Wpo+z09PezevXtMITSZTFy4cMF9PyYmxn07PDycoaEhAHp7e0lMTLzk69tsNi5evMgLL7zgfszlcuF0OgHIycnhyJEjlJeXA3D33Xfz0EMPGcouIvJfUS1WLfYVNcgil8GzSJrNZlasWEFhYeGEcT09PdjtdoaGhtyF2bMgh4eHc/HiRfd9p9OJzWYb89rPPPMM119//YTX/nuCmIrZbKarq4trr712yjHz5s0jLCyMPXv2THpFeEREBPn5+eTn59PR0UFZWRkpKSncdNNNl9y2iIg3qBarFl9pOsVC5F9asWIFDQ0NNDU14XQ6GR4e5vTp0/T29hIfH09KSgqHDx9mZGSElpaWMasbycnJOBwOGhsbGRkZ4dixYzgcDvfzq1ev5uOPP6anpwf4a5Whvr7eUK5Vq1ZRU1ODxWLB5XLR3t7OwMDAmDEmk4lVq1bxwQcfuFdC+vr6aGpqAqChoYHz58/jcrmIiIjAZDLpohAR8UuqxXIlaAVZ5F+Ki4ujpKSEDz/8kIqKCkwmE9dddx1PPvkkAMXFxVRVVbF+/XrS0tK48847sdvtAERGRvLEE09QXV2N0+kkJydnzCHD+++/H4Dy8nL+/PNP5s+fz/Lly8nOzp4214MPPojD4aC8vJyBgQGuvvpqnn/++Qnj1q5dy9GjR9m+fTsDAwPExsayevVqMjMzsVgsvP/++9hsNubOncs999zjviJcRMSfqBbLlaAvChHxksOHD3P+/HmKi4t9HUVEZNZSLRYjtE4vIiIiIuJBDbKIiIiIiAedYiEiIiIi4kEryCIiIiIiHtQgi4iIiIh4UIMsIiIiIuJBDbKIiIiIiAc1yCIiIiIiHv4HR5CR1IZGFIsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "value_freq_graph(X0, X1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Charset**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "X0 = df_exp.loc[y_train == 0, 'char'].value_counts(normalize=True)\n",
    "X1 = df_exp.loc[y_train == 1, 'char'].value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAHwCAYAAAC7apkrAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3XtYVXXe//8newNyUpSQFA+geABS0VTEQ6gTFdlkVn7LbMwOjiNoztxxo1bmaTQ0lUpFJXWyS73rNs0xm7QZPKRoKkLOgOIJTPMIeBgUhA3s/fvDn/uGARVFZIOvx3V1XXuv9Vmf9Vrruli+++zPWsvOYrFYEBERERERAAw1HUBERERExJaoQBYRERERKUUFsoiIiIhIKSqQRURERERKUYEsIiIiIlKKCmQRERERkVJUIIuIiIiIlKICWeQmXn/9dcLCwipcZ2dnx8qVK+9zIhGRB9O1a9f44IMPaNu2Lc7Ozjz00EN0796defPm1XQ0qaPsazqAiIiIyK1ERESwdetWPv30U4KCgsjNzeXnn3/m5MmTNR1N6iiNIItU0aeffkrnzp1xc3OjSZMmDBkyhLNnz1rXb9u2DTs7O77//nt69uyJs7MzXbt25cCBAxw4cIA+ffrg4uJCcHAwBw8erMEjERGxTX/961+Jjo5m0KBBtGrViqCgIF5//XUmTZpkbXPjV7/Y2FiaNWuGi4sLL774Ijk5OdY2KSkpPP3003h5eeHm5kb37t3ZtGlTmX35+vrywQcfEBERgbu7O15eXixYsIDCwkLefvttGjVqRLNmzViwYMF9O365/1Qgi9wDc+bMITU1lXXr1nHy5EmGDBlSrs3777/PjBkzSE5OxtHRkVdeeYWIiAimTp1qXfbGG2/UQHoREdvWtGlTNm3axMWLF2/Zbu/evWzbto1Nmzbx/fff869//Ys333zTuj43N5chQ4awbds2UlJSeOqppxg4cCBHjhwp08/8+fNp27YtycnJjB07lrFjx/L888/TqlUrkpKSGDNmDGPHjtWgRl1mEZEKDR8+3GI0Gi2urq7l/gMsK1asqHC7lJQUC2A5deqUxWKxWLZu3WoBLOvWrbO2Wb16tQWwrFmzxrrsm2++sQCWK1euVO+BiYjUMomJiZaWLVtaDAaDpWPHjpbf//73lr/+9a8Ws9lsbTN8+HCLq6ur5fLly9ZlP/zwgwWwHDly5KZ9d+rUyTJ9+nTrdx8fH8tzzz1n/V5SUmKpX7++5be//W2ZZQ0bNrTMnz//Xh2i2BiNIIvcQo8ePdi/f3+5/0rbtm0bTz31FC1atKB+/fr06dMHgBMnTpRpFxQUZP3cpEkTADp16lRuWVZWVrUci4hIbdW7d28yMjLYsWMHw4cP5/z587z44osMHDgQi8VibRcYGIi7u3uZ7QDS09MByM7OJjIyEn9/fxo2bIibmxsHDhy45fXaYDDQuHHjMtdrg8GAl5eXrtd1mG7SE7kFZ2dn2rRpc9P1J0+eZMCAAQwbNoxJkybh6enJqVOnCAsLw2QylWnr4OBg/WxnZ3fTZWaz+V4egohInWBvb0+vXr3o1asXUVFRrFy5kmHDhrF9+3b69u1bqT5ef/11Tp48yUcffUSrVq1wdnZmyJAht7xew/Xrc0XLdL2uu1Qgi1RBUlIS165d45NPPsHZ2RmA5OTkGk4lIlL3BQQEAGV/dUtPTyc3N5cGDRoAsGvXrjJtt2/fzkcffcTAgQMByMvLIzMzkw4dOtzP6FILaIqFSBW0bdsWOzs75s6dy/Hjx/nrX//KtGnTajqWiEid0rdvXxYvXsy+ffs4ceIEmzdvJjIykoYNG9K/f39rOzs7O1577TXS0tLYvn07o0eP5plnnqFt27YAtG/fnlWrVpGamsr+/ft55ZVXKCkpqanDEhumAlmkCjp16sT8+fOJj48nMDCQOXPm8Mknn9R0LBGROuXpp59m1apVDBgwgPbt2/PGG2/Qtm1bdu7ciaenp7VdcHAwffr04YknnuCpp57ikUce4fPPP7eu//zzzzGbzQQHBzNo0CDCw8Pp3r17TRyS2Dg7S+nZ7SIiIiK10Ouvv86pU6dISEio6ShSB2gEWURERESkFBXIIiIiIiKlaIqFiIiIiEgpGkEWERERESlFBbKIiIiISCl6UUg1OnPmTE1HAMDT05OcnJyajgHYThZbyQG2k8VWcoDtZLmbHN7e3tWURu41W7lG34qt/C3cSm3ICLUjZ23ICLUj580yVvYarQJZREQeSA7rt9R0hNv6N+Bw21Y1qzZkhNqRszZkBNvKWfTcb6qlX02xEBEREREpRSPIIiICwMSJE5k+fXqV+iguLmbx4sUcP34cs9lMaGgozz//PADfffcdW7Zswc7OjhYtWhAZGYmjoyNxcXEcPHgQFxcXAEaPHo2vry9Xr15l0aJFnD9/HgcHByIiImjZsmW5fZ4+fZqFCxdy/PhxhgwZwsCBA6t0DCIiKpBFRASgysUxwO7duykuLmbu3LkUFhbyzjvv0Lt3b+zt7dm4cSMff/wxjo6OxMbGsmvXLvr16wfAsGHDCAkJKdPXunXr8PX1JTo6mtOnT7Ns2TImTZpUbp9ubm688cYbJCUlVTm/iAioQBYRkf/fsGHDWLFiBZcuXeKTTz4hPz8fs9nMiBEjCAgIIDExkXXr1gHQpUsXfve731XYT0FBASUlJZhMJuzt7XFxccFkMmE2mzGZTBiNRkwmE40aNbplnlOnTllHn5s1a0Z2djaXL1+mYcOGZdq5u7vj7u5OSkrKPTgLIiIqkKvV2egRNR0BgLM1HaAUW8liKznAdrLYSg6wnSxngaazl9Z0jPsuMTGRoKAgXnjhBcxmM4WFhVy8eJFVq1Yxa9YsXF1dmT59Onv37iU4OLjMtiEhIezbt4+RI0diMpkYPnw4bm5uADz77LNERETg6OhIUFAQQUFB1u2+/PJL1qxZQ4cOHXj11VdxcHDAx8eHPXv24O/vz7Fjx8jOzubixYvlCmQRkXtNN+mJiEgZfn5+bN26ldWrV3Py5EmcnZ3JyMjgkUceoUGDBhiNRh577DHS09PLbXvs2DEMBgPx8fEsWLCADRs2cP78ea5evUpSUhJxcXHEx8dTUFDA9u3bARg6dCiffPIJMTExXL16lfXr1wMwaNAg8vLyiI6OZuPGjbRq1QqD4e7/2UpISGDChAlMmDDhrvsQkQeDRpBFRKSMwMBApk6dSkpKCvPnz2fgwIE4OztX2Hbv3r18/fXXAIwaNYrExEQ6d+6Mvb097u7utG/fnoyMDOzs7PDy8qJBgwYA9OjRgyNHjhAaGmqdauHg4ED//v3ZsGEDAC4uLkRGRgJgsVgYM2YMXl5ebNq0ic2bNwPw7rvv4uHhUanjCgsLIyws7O5PjIg8MFQgi4hIGdnZ2Xh4eBAWFkZhYSHHjx/nueeeY/ny5eTm5uLm5sbOnTsJDw+ne/fuZaZZpKamkpaWxmOPPUZhYSFHjx7lmWeewWQycfToUQoLC3F0dCQ1NRU/Pz8ALl26RKNGjbBYLCQlJdGiRQsA8vLyqFevHvb29mzevJmAgABcXFwIDw8nPDy8Rs6NiDwYVCCLiEgZBw4cYMOGDRiNRpycnBgzZgyNGjVi6NChTJ06Fbh+k1737t3LbRseHs7ChQuJiorCYrHQv39/fHx8gOvzk8ePH4/RaMTX19c6mjtv3jxyc3MB8PHxYeTIkcD1x7ctWLAAg8FA8+bNGTVqVIV5L1++zIQJE7h27Rp2dnZ8//33xMbGWh8bJyJyp+wsFoulpkPUVcmvDqjpCCJSRXd6k55eNV17ZC9aWdMRRKSKbvYmPb1qWkRE5C5U1ytq76Wb/SNvS2pDRqgdOWtDRqg9OatCT7EQERERESlFBbKIiIiISCmaYiEiIg8k+2/X1HSE27qM7f5DXTxwcE1HEKk2GkEWERERESlFBbKIiIiISCkqkEVEpNaLiYkhLy8PgIkTJ9ZwGhGp7Wx1apOIiEilvfvuu9bP06dPr8EkIlIXqEAWEZEqy8rKYtasWcydOxeAb7/9loKCAtzc3PjHP/6B0WikefPm/OlPfyq37Zo1a0hOTsZkMtGuXTtGjhxpfSPef25bUFDAX/7yFzIyMrCzs2Pw4MGEhIQwevRoYmJiaNCgAcOGDWPFihX3+xSISB2iAllERKrN+vXrWbBgAQ4ODtYpEP8pPDycwYOvPxFh/vz5JCcn061btwq3XbNmDS4uLtZC/OrVq5XOkpCQQEJCAgAzZ86symGJSB2nAllERKpNy5YtmTdvHt27dyc4OLjCNmlpaXz77bcUFhZy9epVWrRoQbdu3SrcNjU1tcwotJubW6WzhIWFERYWVrUDEpEHggpkERGpMqPRiNlstn4vKioCrs8NPnjwIPv27WPt2rXExsYyc+ZMLl++jJ+fH2+++SbLli0jJiYGT09PVq9ejclkuum2AHZ2dvf/AEXkgaICWUREqszd3Z3c3FyuXLmCk5MTKSkpdOrUiZycHDp06IC/vz87d+6koKCA999/37rdjakTDRo0oKCggD179tCjRw/MZnOF23bq1IlNmzbx+uuvA9enWNzJKLKISGWoQBYRkSqzt7fnxRdf5L333sPLywtvb2/MZjPz588nPz8fgGeeeQZXV9cy27m6uvL4448TFRWFl5cXfn5+ADfd9sUXX2Tp0qVERUVhMBgYPHgwPXr0uL8HKyJ1np3FYrHUdIi6KvnVATUdQUSqqOnspXfU3tvbu5qSyL2WtXheTUeo1W68atrT05OcnJwaTnN7tSFnbcgItSPnzTJW9hqtEeRKiomJYezYsbi6ujJx4kQ9Z1NEpJa7UeDZstpQiIjURSqQK0kPoRcRERF5MNSaAlkPoRcRERGR+6HWFMg3o4fQi4jIXVk3p6YT3FaVJ1c8/9/3IobIA6fWF8h6CL2IiIiI3Eu1pkDWQ+hFRO5c6RuMK+M/p7NVlytXrhAbG8uxY8fo168fb731FgCFhYXExsZy/vx5DAYDXbt25dVXXwVg27ZtrFixAg8PD+D6r4OPP/44v/zyC0uWLOHatWsYDAZeeOEFevXqVa35RaRuqzUFsh5CLyJy50rfYGxLHBwcePnllzl58iS//vprmXXPPvssHTp0oLi4mGnTpvHzzz/TpUsXAHr16mUtpm9wdHRkzJgxNG3alIsXLzJhwgSCgoIq/T8FIiL/qdYUyHoIvYhIeevXr8fBwYEBAwawfPlyTpw4weTJk0lNTWXr1q0cPnyYmJgYCgoKiImJoX379hw5cgQPDw/GjRuHo6MjmZmZLFq0CEdHR/z9/a19m0wmli5dSkZGBkajkddee40OHToQExPD0KFD8fHxYdy4cQQHBzN48GC++uorGjduzKOPPsonn3xCfn4+ZrOZESNGEBAQUCa3k5MT/v7+nDt3rszyevXq0aFDB+D6db9Vq1ZcuHDhlueg9HNNPTw8rAMqKpBF5G7VmgIZYMCAAQwYcOcv3xgyZAhDhgwpt/zPf/5zuWVOTk6MGTOm3PK4uDjrZz3BQkRsRUBAAN999x0DBgwgMzOToqIiiouLOXToEAEBARw+fNja9uzZs/zxj39k1KhRxMbGsnv3bkJDQ1m4cCFvvvkmgYGBZa5vP/zwAwBz587l9OnTTJ8+nU8//ZSAgADS09Np3LgxRqPRuo/Dhw8TGhpKYmIiQUFBvPDCC5jNZgoLC+/q2PLy8khOTi5z3d+zZw/p6ek0bdqU4cOH4+npWWabY8eOUVxczMMPP1yuP91ILSKVZajpACIicvdat25NZmYm165dw8HBgXbt2pGZmWktkEvz8vLC19fXul12djb5+fnk5eURGBgIQGhoqLX9oUOHrN+bNWtG48aNOXv2LP7+/qSnp3Po0CG6dOlCQUEBhYWFZGVl4e3tjZ+fH1u3bmX16tWcPHkSZ2fnOz6ukpISPv30U55++mlrsdu1a1fi4uKYM2cOHTt2LDNwAXDp0iXmz59PREQEBkP5f97CwsKYOXOmimMRua1aNYIsIiJl2dvb07hxY7Zu3Uq7du3w8fEhLS2Nc+fO0axZszJtHRwcrJ8NBgMmkwmLxXLTG5MtFkuFy9u0aUNmZiYPP/wwnTp14sqVK2zevJnWrVsDEBgYyNSpU0lJSWH+/PkMHDgQZ2dnvv76awBGjRplne52M/Hx8TRp0oRnnnnGuqx+/frWz2FhYaxatcr6PT8/n5kzZzJkyBDatWt3y75FRG5HI8giIrVcQEAAGzZsICAgAH9/f/7xj3/g6+tbqSfyuLq64uLiwqFDhwDYsWOHdV1gYKD1+5kzZ8jJycHb2xt7e3seeughfvrpJ9q2bVtm/wDZ2dm4u7sTFhbGb37zG44fP05wcDCzZ89m9uzZty2Ov/rqK/Lz8603S99w6dIl6+d9+/bRvHlzAIqLi5kzZw6hoaH07Nnz9idMROQ2NIIsIlLLBQQEsG7dOtq1a4eTkxOOjo7lplfcSmRkpPUmvaCgIOvyJ598kiVLlhAVFYXRaCQyMtI6Cu3v709aWhr16tXD39+fCxcuWG/wO3DgABs2bMBoNN70vg6A0aNHk5+fT3FxMUlJSUycOBFnZ2e++eYbmjVrxvjx44H/e5zbxo0b2bdvH0ajETc3NyIjIwHYtWsX6enpXLlyhW3btln7vjGdRETkTtlZbvYbmlRZ8qt3fkOhiNiWprOX3lH70k9UENt2Ju6dmo5Q/e7Dm/Q8PT3JyanyO/+qXW3IWRsyQu3IebOMlb1GawRZREQeTLXgNcy1oRARqYs0B1lEREREpBQVyCIiIiIipWiKhYiIPJBy10XcdF2D5xfdxyQiYms0giwiIiIiUooKZBERERGRUlQgi4hIjcvKyiIqKqrc8sWLF3Pq1Klbbrtt2zbeeustoqOjiY6OZvPmzdUVU0QeEJqDLCIiNmvUqFGVaterVy/eeuutak4jIg8KFcgiImITSkpKWLBgAb/88gtNmzZlzJgxxMTEMGzYMPz8/NiyZQvr16+nUaNGNGnSBAcHBxXFIlItVCCLiIhNOHPmDKNGjcLf35+FCxfyww8/WNddvHiRtWvXMmvWLJycnJg2bRo+Pj7W9Xv27CE9PZ2mTZsyfPhwPD09y/WfkJBAQkICADNnzqz+AxKRWksFsoiI2ISHHnoIf39/AEJDQ/n++++t644dO0ZAQABubm4AhISEcPbsWQC6du1K7969cXBw4O9//ztxcXFMnjy5XP9hYWGEhYXdhyMRkdpON+mJiIhNsLOzu+X3m6lfvz4ODg7A9SI4MzPznmcTkQeLCmQREbEJOTk5HDlyBIDExETraDJAmzZtSE9P5+rVq5SUlLBnzx7rukuXLlk/79u3j+bNm9+/0CJSJ2mKhYiI2IRmzZqxbds2PvvsM5o0acKTTz5JcnIyAB4eHjz//PO8//77NGrUiObNm+Pi4gLAxo0b2bdvH0ajETc3NyIjI2vyMESkDrCzWCyWmg5RV505c6amIwDg6elJTk5OTccAbCeLreQA28liKznAdrLcTQ5vb+9qSiMFBQU4OTlRUlLC7Nmz+c1vfkNwcPBd92cr1+hbsZW/hVupDRmhduSsDRmhduS8WcbKXqM1giwiIrXC6tWrSU1NpaioiE6dOtG9e/eajiQidZQKZBERqRVee+21mo4gIg8IFcgiIvJAOvHdSOtnn99+VoNJRMTW6CkWIiIiIiKlqEAWERERESlFBbKIiFSb1atX8+233950fW5uLu+99x7jxo0jPT39jvvftm0by5YtA2Dv3r2cOnXqrrOKiNygAllERGpMamoq3t7efPTRRwQEBFSpr6SkJBXIInJP6CY9ERG5p7755ht+/PFHPD09qV+/Pq1bt+bcuXMsW7aM3Nxc6tWrxx/+8AeKiopYuXIlJpOJ6OhoZsyYwRdffEFGRgYmk4mQkBBeeuklAEaPHk1MTAwNGjQgIyODFStWMGXKFOs+Dx8+zL59+zh48CBr164lKiqKJk2a1NAZEJHaTgWyiIjcM5mZmezcuZOPPvqIkpISxo8fT+vWrfnss8/4/e9/T9OmTTl69ChLly5l8uTJvPzyy2RkZPDWW28B8Morr+Dm5obZbGbatGmcOHECHx+f2+63ffv2dOvWja5duxISElJhm4SEBBISEgCYOXPmvTtoEalzVCCLiMg9k56eTnBwMPXq1QOgW7duFBUVcfjwYWJjY63tiouLK9x+165dbN68mZKSEi5dusSpU6cqVSBXRlhYGGFhYfekLxGp21Qgi4jIPWVnZ1fmu8ViwdXVldmzZ99yu6ysLDZs2EBMTAxubm7ExcVRVFQEgMFgwGKxAFiXiYhUFxXI1ej1L36q6Qh10vLhPWs6gojcREBAAAsXLmTQoEGUlJSQnJxMWFgYXl5e/PTTT/Ts2ROLxcKJEyfw9fUts21+fj5OTk64uLhw+fJl9u/fzyOPPAKAl5cXmZmZdOnShd27d1e4b2dnZ65du1bdhygiDwAVyCIics+0bt2aXr16ER0dTePGjfH39wdg7NixLFmyhG+++Ybi4mJ69+5drkD29fXF19eXqKgovLy8aN++vXXd4MGDWbx4MevWraNNmzYV7rtXr17Ex8ezceNG3nnnHd2kJyJ3zc5y4zcrueeejFlb0xHqpHsxguzp6UlOTs49SFN1tpLFVnKA7WS5mxze3t7VlEbutZ8++631s62+atpW/hZupTZkhNqRszZkhNqR82YZK3uN1giyiIg8kGy1KBaRmqcXhYiIiIiIlKICWURERESkFE2xEBGRB1LqxhHWzx2fXlqDSUTE1mgEWURERESkFI0gi4hItZkyZQrDhg3Dz8+vzPIdO3awfv16AJycnBgxYgS+vr6YTCYmT55McXExJSUlhISE8NJLLwGwadMm/va3v3H+/HmWLl1KgwYNbtmXiMjdUoEsIiL3nZeXF1OmTMHNzY2ff/6Zzz77jA8//BAHBwcmT56Mk5MTxcXFTJo0ic6dO9OuXTvat2/Po48+ytSpUyvVl4jI3VKBLCIi98SaNWtITEzkoYceon79+rRu3Rq4PsL7+eefc+3aNSIiImjTpk2Zl4C0bduWCxcuANdfU+3k5ARASUkJJSUl1ldXt2rVqsL93qwvEZG7pQJZRESqLCMjgz179vDRRx9RUlLC+PHjrQVyQUEB06dP5+DBgyxatIi5c+eW2XbLli106dLF+t1sNjN+/HjOnTvHU089Rdu2bSud4z/7Ki0hIYGEhAQAZs6ceaeHKCIPEBXIIiJSZYcOHaJ79+44OjoC0LVrV+u6Pn36ABAYGEh+fj55eXm4uroCkJaWxtatW5k2bZq1vcFgYPbs2eTl5TFnzhxOnjxJy5Ytb5uhor5KCwsLIyws7K6PUUQeHHqKhYiIVJnFYql02xtTJk6cOEF8fDzR0dHUr1+/XDtXV1cCAwPZv3//bfu8XV8iIndCBbKIiFSZv78/ycnJmEwmCgoKSElJsa7btWsXcH2U2cXFBRcXF3JycpgzZw5jxozB29vb2jY3N5e8vDwATCYTqampNGvW7Jb7vllfIiJ3S1MsRESkytq0aUPXrl2Jjo6mcePG+Pn54eLiAoCbmxsTJ0603qQH12/ou3r1KkuXXn9Bh9FoZObMmVy6dIm4uDjMZjMWi4WePXtap2t8//33fPvtt1y+fJno6Gi6dOnCqFGjbtqXiMjdsrPcye9ickeejFlb0xHqpOXDe1a5D09PT3Jycu5BmqqzlSy2kgNsJ8vd5HiQRzALCgpwcnKisLCQyZMnM3LkSOuNerboh2UDrJ9t9U16tvK3cCu1ISPUjpy1ISPUjpw3y1jZa7RGkEVE5J6Ij4/n1KlTFBUV0bdvX5sujsF2i2IRqXkqkEVE5J744x//WNMRRETuCd2kJyIiIiJSigpkERF5IO364U12/fBmTccQERukAllEREREpJT7Ogd54sSJTJ8+vUp9FBcXs3jxYo4fP47ZbCY0NJTnn38egO+++44tW7ZgZ2dHixYtiIyMxNHRkbi4OA4ePGh95NDo0aPx9fXl6tWrLFq0iPPnz+Pg4EBERESFb2s6ffo0Cxcu5Pjx4wwZMoSBAwdW6RhERERExHbd1wK5qsUxwO7duykuLmbu3LkUFhbyzjvv0Lt3b+zt7dm4cSMff/wxjo6OxMbGsmvXLvr16wfAsGHDCAkJKdPXunXr8PX1JTo6mtOnT7Ns2TImTZpUbp9ubm688cYbJCUlVTm/iIiIiNi2+zrFYtiwYQBcunSJyZMnEx0dTVRUFOnp6QAkJiYSFRVFVFQUK1euvGk/BQUFlJSUYDKZsLe3t44Mm81mTCaTdV2jRo1umefUqVN07NgRgGbNmpGdnc3ly5fLtXN3d6dNmzYYjca7Om4RkXtt9erVfPvttzddn5uby3vvvce4ceOs19g7sW3bNpYtWwbA3r17OXXq1B33ceDAgfv6wo6srCwSExPv2/5EpO6qkce8JSYmEhQUxAsvvIDZbKawsJCLFy+yatUqZs2ahaurK9OnT2fv3r0EBweX2TYkJIR9+/YxcuRITCYTw4cPx83NDYBnn32WiIgIHB0dCQoKIigoyLrdl19+yZo1a+jQoQOvvvoqDg4O+Pj4sGfPHvz9/Tl27BjZ2dlcvHiRhg0b3tVxJSQkkJCQAKC3OIlIjUpNTcXb25sxY8ZUua+kpCS6du1K8+bN70GyyispKbmjgYns7GwSExPp06dPNaYSkQdBjRTIfn5+LFq0iOLiYoKDg/H19SUtLY1HHnmEBg0aAPDYY4+Rnp5erkA+duwYBoOB+Ph48vLymDRpEh07dsTV1ZWkpCTi4uJwcXEhNjaW7du3ExoaytChQ2nYsCHFxcXEx8ezfv16Bg8ezKBBg1i+fDnR0dG0bNmSVq1aYTDc/aB6WFgYYWFhVTo3IiI388033/Djjz/i6elJ/fr1ad26NefOnWPZsmXk5uZSr149/vCHP1BUVMTKlSsxmUxER0czY8YMvvjiCzIyMjCZTISEhPDSSy8B1+/JiImJoUGDBmRkZLBixQqmTJli3efhw4fZt28fBw8eZO3atURFRdGkSZNy2c6dO8eSJUvIzc3FYDDwX//1X8D8feeQAAAgAElEQVT1X/zmzp3Lr7/+SuvWrXn77bexs7NjzZo1JCcnYzKZaNeuHSNHjsTOzo4pU6bQrl07Dh8+TLdu3fD09GTNmjUYDAZcXFyYOnUqZrOZVatWcfDgQYqKinjqqad44okn+J//+R9OnTpFdHQ0ffv25be//W2ZjBrEEJHKqpECOTAwkKlTp5KSksL8+fMZOHAgzs7OFbbdu3cvX3/9NQCjRo0iMTGRzp07Y29vj7u7O+3btycjIwM7Ozu8vLysBXaPHj04cuQIoaGh1qkWDg4O9O/fnw0bNgDg4uJCZGQkABaLhTFjxuDl5cWmTZvYvHkzAO+++y4eHh7Vej5ERG4nMzOTnTt38tFHH1FSUsL48eNp3bo1n332Gb///e9p2rQpR48eZenSpUyePJmXX36ZjIwM3nrrLQBeeeUV3NzcMJvNTJs2jRMnTuDj43Pb/bZv355u3brRtWvXcvdxlDZv3jwGDRpEcHAwJpMJi8XChQsXOH78OLGxsTRq1IgPPviAw4cP4+/vT3h4OIMHDwZg/vz5JCcn061bNwDy8/OZOnUqAFFRUbz//vt4eHiQl5cHwJYtW3BxcSEmJoaioiI++OADgoKCGDp0KBs2bGDChAkVZtQghohUVo0UyNnZ2Xh4eBAWFkZhYSHHjx/nueeeY/ny5eTm5uLm5sbOnTsJDw+ne/fuZUaRU1NTSUtL47HHHqOwsJCjR4/yzDPPYDKZOHr0KIWFhTg6OpKamoqfnx9wfc5zo0aNsFgsJCUl0aJFCwDy8vKoV68e9vb2bN68mYCAAFxcXAgPDyc8PLwmTo2ISIVu/KJWr149ALp160ZRURGHDx8mNjbW2q64uLjC7Xft2sXmzZspKSnh0qVLnDp1qlIFcmVcu3aNixcvWq/Vjo6O1nVt2rThoYceAsDX15esrCz8/f1JS0vj22+/pbCwkKtXr9KiRQtrgdyrVy/r9u3btycuLo6ePXvSo0cPAP75z39y8uRJdu/eDVwvqM+ePYu9vV4OKyL3Ro1cTQ4cOMCGDRswGo04OTkxZswYGjVqxNChQ62jBl26dKF79+7ltg0PD2fhwoVERUVhsVjo37+/9SIfEhLC+PHjMRqN+Pr6WkcK5s2bR25uLgA+Pj6MHDkSuP74tgULFmAwGGjevDmjRo2qMO/ly5eZMGEC165dw87Oju+//57Y2FjrzYEiIveDnZ1dme8WiwVXV1dmz559y+2ysrLYsGEDMTExuLm5ERcXR1FREQAGgwGLxQJgXXanbmxfEQcHB+tng8FgvZl62bJlxMTE4OnpyerVqzGZTNZ2N/4nAGDkyJEcPXqUlJQUxo0bx0cffYTFYuGNN96gc+fOZfZ14MCBu8ovIvKf7muBvGLFCgD69etnffxaaX369LntzRVOTk688847Fa576aWXrPPqSps8eXKF7du1a8e8efNukxoaNmzI4sWLb9tORKS6BAQEsHDhQgYNGkRJSQnJycmEhYXh5eXFTz/9RM+ePbFYLJw4cQJfX98y2+bn5+Pk5ISLiwuXL19m//79PPLIIwB4eXmRmZlJly5drCOy/8nZ2Zlr167dNJuLiwsPPfSQ9cbqoqIizGbzTdvfKMQbNGhAQUEBe/bssY4O/6dz587Rtm1b2rZtS3JyMhcuXKBz5878/e9/p0OHDtjb23PmzBk8PDxum1NEpLL0e5SISC3QunVrevXqRXR0NI0bN8bf3x+AsWPHsmTJEr755huKi4vp3bt3uQLZ19cXX19foqKi8PLyon379tZ1gwcPZvHixaxbt442bdpUuO9evXoRHx/Pxo0beeeddyq8SW/MmDF89tlnrF69GqPReNOBDABXV1cef/xxa54b0+EqsnLlSs6ePQtAhw4d8PHxoWXLlmRlZTF+/HjgeqF942Zro9F405v0REQqy85yq9/GpEqejFlb0xHqpOXDe1a5D09PT3Jycu5BmqqzlSy2kgNsJ8vd5PD29q6mNHKvnTlzpqYj3Jat/C3cSm3ICLUjZ23ICLUj580yVvYafV9fFCIiIiIiYus0xUJERCpt6dKlHD58uMyyAQMG0L9//xpKdPd++MfrADz1xPKajCEiNkgFsoiIVNqIESNqOoKISLXTFAsRERERkVI0giwiIgBMnDiR6dOnV6mP4uJiFi9ezPHjxzGbzYSGhvL8888D8N1337Flyxbs7Oxo0aIFkZGRODo6EhcXx8GDB63Plh89ejS+vr5cvXqVRYsWcf78eRwcHIiIiKBly5bl9rljxw7Wr18PXH8U6IgRI8o9yUNE5E6oQBYREYAqF8cAu3fvpri4mLlz51JYWMg777xD7969sbe3Z+PGjXz88cc4OjoSGxvLrl27rM/EHzZsWLlXWa9btw5fX1+io6M5ffo0y5YtY9KkSeX26eXlxZQpU3Bzc+Pnn3/ms88+48MPP6zysYjIg0sFsoiIANeL1BUrVnDp0iU++eQT8vPzMZvNjBgxgoCAABITE1m3bh1w/W2nv/vd7yrsp6CggJKSEkwmE/b29ri4uGAymaxv0TMajZhMJho1anTLPKdOnbKOPjdr1ozs7GwuX75Mw4YNy7Qr/Vzntm3bcuHChaqcBhERFcgiIlJWYmIiQUFBvPDCC5jNZgoLC7l48SKrVq1i1qxZuLq6Mn36dOub80oLCQlh3759jBw5EpPJxPDhw3FzcwPg2WefJSIiAkdHR4KCgggKCrJu9+WXX7JmzRo6dOjAq6++ioODAz4+PuzZswd/f3+OHTtGdnY2Fy9eLFcgl7Zlyxa6dOlS4bqEhAQSEhIAmDlzZlVPk4jUYSqQRUSkDD8/PxYtWkRxcTHBwcH4+vqSlpbGI488QoMGDQB47LHHSE9PL1cgHzt2DIPBQHx8PHl5eUyaNImOHTvi6upKUlIScXFxuLi4EBsby/bt2wkNDWXo0KE0bNiQ4uJi4uPjWb9+PYMHD2bQoEEsX77c+pa8Vq1aYTDc/N7ytLQ0tm7dyrRp0ypcHxYWRlhY2L07USJSZ6lAFhGRMgIDA5k6dSopKSnMnz+fgQMH4uzsXGHbvXv38vXXXwMwatQoEhMT6dy5M/b29ri7u9O+fXsyMjKws7PDy8vLWmD36NGDI0eOEBoaap1q4eDgQP/+/dmwYQMALi4uREZGAmCxWBgzZgxeXl5s2rSJzZs3A/Duu+/i4eHBiRMniI+P591336V+/frVen5EpO5TgSwiImVkZ2fj4eFBWFgYhYWFHD9+nOeee47ly5eTm5uLm5sbO3fuJDw8nO7du5cZRU5NTSUtLY3HHnuMwsJCjh49yjPPPIPJZOLo0aMUFhbi6OhIamoqfn5+AFy6dIlGjRphsVhISkqiRYsWAOTl5VGvXj3s7e3ZvHkzAQEBuLi4EB4eTnh4uHWfOTk5zJkzhzFjxuhV3yJyT6hAFhGRMg4cOMCGDRswGo04OTkxZswYGjVqxNChQ5k6dSpw/Sa97t27l9s2PDychQsXEhUVhcVioX///vj4+ADX5yePHz8eo9GIr6+vdbrDvHnzyM3NBcDHx4eRI0cCcPr0aRYsWIDBYKB58+aMGjWqwrxr1qzh6tWrLF26FACj0ag5xiJSJXYWi8VS0yHqqidj1tZ0hDpp+fCeVe7D09OTnJyce5Cm6mwli63kANvJcjc5NIJZe3z+xZOAbb9q2lb+Fm6lNmSE2pGzNmSE2pHzZhkre43WCLKIiDyQbLkwFpGapVdNi4iIiIiUogJZRERERKQUTbGoRvdiruy9YEtzhWwpi4g82NZsfYPB/T+v6RgiYoM0giwiIiIiUopGkEVExOZs27aNTp064eHhAUB6ejpLlizBaDQyY8YMHB0drW1zcnKIi4sjLy8Ps9nM0KFDefTRR2squojUASqQRUTE5mzbto0WLVpYC+QdO3bw7LPP0r9//3Jt165dS8+ePXnyySc5deoUMTExKpBFpEpUIIuISI3Jyspi1qxZzJ07F4Bvv/2WH3/8kaysLObNm4ejoyO/+c1v+Omnn/jnP/9JamoqY8eOLdOHnZ0d+fn5AOTn51tfXS0icrdUIIuIiE3p0aMHBw8eZNiwYdbXUWdmZtK1a1dCQkLKtf9//+//MX36dDZt2kRhYSEffPDB/Y4sInWMCmQREanVdu7cSb9+/Xj22Wc5cuQI8+fPZ+7cuRgMZe9DT0hIICEhAUCvohaRW1KBLCIiNcZoNGI2m63fi4qKbrvNl19+SUpKCgCzZ89my5YtvPfeewC0a9eOoqIirly5gru7e5ntwsLCCAsLu4fpRaSuUoEsIiI1xt3dndzcXK5cuYKTkxMpKSkEBQXh5OTEtWvXKtzmlVde4ZVXXrF+9/T0JC0tjX79+nHq1CmKiopo0KDB/ToEEamDVCCLiEiNsbe358UXX+S9997Dy8sLb29vAPr168eSJUtwdHRkxowZt+zjtddeIz4+nr/97W8AREZGYmdnV+3ZRaTuUoEsIiI1asCAAQwYMKDc8tI35I0ePfqm2zdv3pw///nP1ZJNRB5MepOeiIg8kPSaaRG5GRXIIiIiIiKlaIpFNfrh27O3bfPUwKb3IYmIiIiIVJZGkEVE5IH0WeIbNR1BRGyUCmQRERERkVJUIIuIiIiIlKI5yCIiNsJkMjF58mSKi4spKSkhJCSEl156qUybAwcOsGHDBiZMmHDb/jIyMvjxxx958803K51h9erVODk5MXDgwDvOfyd++uknvv76a06fPs2HH36In58fAP/6179YtWoVxcXF2NvbM2zYMDp06ADAlClTuHTpEo6OjgBMnDgRd3d3vvvuOzZv3ozRaKRBgwZERETQuHHjas0vInWbCmQRERvh4ODA5MmTcXJyori4mEmTJtG5c2fatWt3V/35+flZC09b06JFC/77v/+bzz77rMzy+vXrM378eDw8PDh58iQzZswgPj7eun7s2LHljsnX15eZM2dSr149/v73v7Ny5Ur+67/+674ch4jUTSqQRURshJ2dHU5OTgCUlJRQUlKCnZ0d+/fvZ/ny5dSvX59WrVpZ20dFRTFt2jRcXFx46623GD58OH379mX+/Pn07dsXo9FoHW1evXo1OTk5ZGVlkZOTU+blHN988w0//vgjnp6e1K9fn9atWwPwyy+/sGTJEgoLC3n44YeJiIigpKSEDz/8kFmzZvHLL78wbtw4Fi5ciKenJ2+//TZz5swhJSWFNWvWYDAYcHFxYerUqeWOtXnz5hWeg9LH16JFC4qKiigqKsLBweGm5+3GCDNA27Zt2bFjxx2cdRGR8lQgi4jYELPZzPjx4zl37hxPPfUUPj4+/PGPf2TSpEk0adKEjz/+2Nq2ffv2HD58GE9PTx5++GHS09Pp27cvR48e5fe//z0ZGRll+j5z5gyTJ0/m2rVr/OlPf+LJJ5/k5MmT7Ny5k48++oiSkhLGjx9vLZAXLFjAm2++SWBgIP/7v//LmjVreP311ykqKiI/P59Dhw7h5+dHeno6/v7+NGjQgHr16rFmzRref/99PDw8yMvLu+tzsWfPHlq1alWmOF64cCEGg4EePXrw4osvlnul9JYtW+jcuXOF/SUkJJCQkADAzJkz7zqXiNR9KpBFRGyIwWBg9uzZ5OXlMWfOHDIzM/Hy8qJp0+vPTA8NDbUWeQEBARw8eJDGjRvzxBNPsHnzZi5evIibm5t1JLq0Rx99FAcHBxwcHHB3d+ff//436enpBAcHU69ePQC6desGQH5+Pnl5eQQGBgLQt29fa3Herl07Dh8+zMGDB3n++efZv38/FouFgIAA4HrhHhcXR8+ePenRo8ddnYdff/2VVatW8f7771uXjR07Fg8PD65du8bcuXPZvn07ffv2ta7fvn07mZmZTJkypcI+w8LCCAsLu6s8IvJg0VMsRERskKurK4GBgRw5cuSmbQICAjh06BDp6ek88sgj1K9fn927d+Pv719he3v7/xsTMRgMlJSUAJQbhb2dgIAA0tPTycnJoVu3bpw4cYJDhw5ZC+SRI0cyZMgQLly4wLhx47hy5QoLFy4kOjqamJiY2/Z/4cIF5syZw+jRo2nSpIl1uYeHBwDOzs706dOHY8eOWdf961//Yt26dYwbN+6W0zFERCpDBbKIiI3Izc21TkkwmUykpqbSrFkzsrKyOHfuHACJiYnW9p6enly5coVz587x8MMP4+/vz4YNG6yFamUEBASwd+9eTCYT165dIzk5GQAXFxfc3NxIT08Hro/O3ug3ICCAHTt20KRJEwwGA25ubvz888+0b98egHPnztG2bVtefvll6tevz4ULF4iMjGT27Nm8++67t8yTl5fHzJkzeeWVV8oU+iUlJeTm5gJQXFxMcnIyLVq0AOD48eMsWbKEcePG4e7uXuljFxG5GU2xEBGxEZcuXSIuLg6z2YzFYqFnz5507doVo9HIzJkzqV+/Pv7+/vz666/Wbdq0aYPZbAauF65ffvnlTUeQK9K6dWt69epFdHQ0jRs3LrPt6NGjrTfpeXl5ERkZCYCXlxeAdfpF+/btuXDhAm5ubgCsXLmSs2fPAtdvoPPx8Sm337179/KXv/yF3NxcZs6cia+vL++//z6bNm3i3LlzrF27lrVr1wLXH+dWr149ZsyYQUlJCWazmY4dO1qnS6xcuZKCggJiY2OB6//jMH78+EqfAxGR/2RnsVgsNR2irvp8cfJt2zw1sGm15/D09CQnJ6fa91MZtpLFVnKA7WSxlRxgO1nuJoe3t3c1pZF7bcrqpxjZ5/OajnFLtvK3cCu1ISPUjpy1ISPUjpw3y1jZa7SmWIiIyAPJ1otjEak5KpBFREREREpRgSwiIiIiUooKZBEReSC9v+fNmo4gIjZKBbKIiIiISCl6zJuISB0WExPD2LFjcXV1rVT7rKwsZs2axdy5c6s115UrV4iNjeXYsWP069ePt956C4DCwkJiY2M5f/48BoOBrl278uqrrwKwbds2VqxYYX1hSHh4OI8//jgAM2bM4OjRo/j7+zNhwoRqzS4idZ8KZBGROux2L+aoKQ4ODrz88sucPHmyzHOdAZ599lk6dOhAcXEx06ZN4+eff6ZLly4A9OrVy1pMlzZw4EAKCwutr+EWEakKFcgiIrXY+vXrcXBwYMCAASxfvpwTJ04wefJkUlNT2bp1K4cPHyYmJoaCggJiYmJo3749R44cwcPDg3HjxuHo6EhmZiaLFi3C0dGxzItCTCYTS5cuJSMjA6PRyGuvvUaHDh2IiYlh6NCh+Pj4MG7cOIKDgxk8eDBfffUVjRs35tFHH+WTTz4hPz8fs9nMiBEjyr3dz8nJCX9/f+sbAm+oV68eHTp0AK6/GrtVq1ZcuHDhtuehY8eOHDhw4B6cURGRaiqQJ06cyPTp06vUR3FxMYsXL+b48eOYzWZCQ0N5/vnnAfjuu+/YsmULdnZ2tGjRgsjISBwdHYmLi+PgwYO4uLgA198C5evry9WrV1m0aBHnz5/HwcGBiIgIWrZsWW6fp0+fZuHChRw/fpwhQ4YwcOBA67rRo0fj5OSEwWCwvtVKRKSmBQQE8N133zFgwAAyMzMpKiqiuLiYQ4cOERAQwOHDh61tz549yx//+EdGjRpFbGwsu3fvJjQ0lIULF/Lmm28SGBjIihUrrO1/+OEHAObOncvp06eZPn06n376KQEBAaSnp9O4cWOMRqN1H4cPHyY0NJTExESCgoJ44YUXMJvNFBYW3tWx5eXlkZyczIABA6zL9uzZQ3p6Ok2bNmX48OF4enpWur+EhATrCLOu4SJyK9VSIFe1OAbYvXs3xcXFzJ07l8LCQt555x169+6Nvb09Gzdu5OOPP8bR0ZHY2Fh27dpFv379ABg2bBghISFl+lq3bh2+vr5ER0dz+vRpli1bxqRJk8rt083NjTfeeIOkpKQKM02ePJkGDRpU+dhERO6V1q1bk5mZybVr13BwcKBVq1ZkZmZy6NAh3njjDf76179a23p5eeHr62vdLjs7m/z8fPLy8qyvjQ4NDWX//v0AHDp0iKeffhqAZs2a0bhxY86ePYu/vz8bN27Ey8uLLl26kJqaSmFhIVlZWXh7e3P58mUWLVpEcXExwcHB1n3eiZKSEj799FOefvppHn74YQC6du1K7969cXBw4O9//ztxcXFMnjy50n2GhYVZX08tInIr1VIgDxs2jBUrVnDp0qUKf2ZLTExk3bp1AHTp0oXf/e53FfZTUFBASUkJJpMJe3t7XFxcMJlMmM1mTCYTRqMRk8lEo0aNbpnn1KlT1tHnZs2akZ2dzeXLl2nYsGGZdu7u7ri7u5OSknIPzoKISPWzt7encePGbN26lXbt2uHj40NaWhrnzp2jWbNmZdo6ODhYPxsMBkwmExaLBTs7uwr7tlgsFS5v06YNmZmZPPzww3Tq1IkrV66wefNmWrduDUBgYCBTp04lJSWF+fPnM3DgQJydnfn6668BGDVqFH5+frc8rvj4eJo0acIzzzxjXVa/fn3r57CwMFatWnXLPkRE7la1zkGu6Ge2ixcvsmrVKmbNmoWrqyvTp09n7969BAcHl9k2JCSEffv2MXLkSEwmE8OHD8fNzQ24fgNHREQEjo6OBAUFERQUZN3uyy+/ZM2aNXTo0IFXX30VBwcHfHx82LNnD/7+/hw7dozs7GwuXrxYrkC+nRkzZgDwxBNPVDgKoZ/vRKQmBAQEsGHDBuv0sS+++ILWrVvftPAtzdXVFRcXFw4dOoS/vz87duywrgsMDGTHjh106NCBM2fOkJOTg7e3N/b29jz00EP89NNPvPjii+Tm5rJixQqeffZZALKzs/Hw8CAsLIzCwkKOHz/O66+/Xu46fzNfffUV+fn5jBo1qszyS5cuWQdE9u3bR/PmzSt7ikRE7ki1Fsh+fn7lfmZLS0vjkUcesU5VeOyxx0hPTy934Tx27BgGg4H4+Hjy8vKYNGkSHTt2xNXVlaSkJOLi4nBxcSE2Npbt27cTGhrK0KFDadiwIcXFxcTHx7N+/XoGDx7MoEGDWL58OdHR0bRs2ZJWrVphMNzZI6D//Oc/4+Hhwb///W+mT5+Ot7e39SfJG/TznYjUhICAANatW0e7du1wcnLC0dGx3E1xtxIZGWm9Sa/0gMOTTz7JkiVLiIqKwmg0EhkZaR2F9vf3Jy0tjXr16uHv78+FCxesN/gdOHCADRs2YDQacXJyYsyYMRXud/To0eTn51NcXExSUhITJ07E2dmZb775hmbNmjF+/Hjg/x7ntnHjRvbt24fRaMTNzY3IyEhrX5MmTeL06dMUFBQwatQoRo0aRefOne/4XIqIANhZbvYbWhXcmGIBcPHiRVJSUti4caP1Z7a9e/daL5hbtmzh119/JSAgoMzPbzd+LgwNDQVg4cKFdO7cGTs7O/bv309ERAQAP/74I0ePHmXEiBFlMty4QP/n8zAtFgtjxoxh9uzZbN++nc2bNwPXH4V049maq1evxsnJqcxNeqXdbv0Nny9Ovu25empg09u2qSpPT09ycnKqfT+VYStZbCUH2E4WW8kBtpPlbnJ4e3tXUxq5195YF86MHn+p6Ri3ZCt/C7dSGzJC7chZGzJC7ch5s4yVvUZX6whyRT+zPffccyxfvpzc3Fzc3NzYuXMn4eHhdO/evcwocmpqKmlpaTz22GMUFhZy9OhRnnnmGUwmE0ePHqWwsBBHR0dSU1Otc9lu/PxmsVhISkqiRYsWwPU7oevVq4e9vT2bN28mICAAFxcXwsPDCQ8Pv+1xFBQUYLFYcHZ2pqCggH/9618MHjy4ek6aiIjcF7ZeHItIzanWArmin9kaNWrE0KFDmTp1KnD9Jr3u3buX2zY8PJyFCxcSFRWFxWKhf//++Pj4ANfnJ48fPx6j0Yivr691WsO8efPIzc0FwMfHh5EjRwLXH9+2YMECDAYDzZs3Lzev7YbLly8zYcIErl27hp2dHd9//z2xsbFcuXKFOXPmANfvrO7Tp49+uhMRERGpo6plioVcpykW5dlKFlvJAbaTxVZygO1k0RSLuu3MmTM1HeG2bOVv4VZqQ0aoHTlrQ0aoHTmrOsXizu5UExERERGp41Qgi4iIiIiUogJZRERERKSUar1JT0RE7l5WVhazZs1i7ty51bqfbdu2sWLFCuujLm88dxggJyeHxYsXc+HCBeD6IzG9vLzIysrik08+4erVq7Rq1Yq3334be3t7Dh48yBdffMGJEyf405/+REhISIX7vFW7lStX8vPPPwPw4osv0qtXL+D6041WrlyJ2WzGycmJ0aNH06RJE5YvX86BAwcAMJlM/Pvf/2b58uXVcq5E5MGgAllEROjVqxdvvfVWueULFizg/2PvTqOivNL9739roGTUiIgKKiCKVQjiiJhEWtN1bGJak6j/DA7RdFwOaGvSBMXTDkHlgCI4BY0xWW238cRoom1It5rGMcQJtZMGBQ1gnBAExIAgVRRVzwse60AYJA4R5Pq8qmHf+/7dtVaVl5t97z169Gh69+5NeXm5dXe+Tz75hBdeeIFnnnmGDz/8kAMHDjB8+HBcXFwIDQ0lMTGxwfPV1+7MmTNcvHiRFStWUFFRwXvvvUefPn2wt7fno48+Ijw8nM6dO7Nv3z6++OILZs6cyeTJk63H79mzh4sXLz74ByKEaNGkQBZCiGYgLy+PuLg4Jk+ezJYtWzCZTFgsFsLCwujUqe7VcA4fPkxiYiIKhYKuXbvyxz/+8Red8+rVq1RWVtK7d28AbG1tgaoNl86ePcucOXMAGDp0KDt27GD48OG4uroC3HOb6/raXb16FV9fX1QqFSqVCg8PD7777jvrKPKdO3cAKCsrs247Xd23337LK6+88ouuUwghfk4KZCGEaOJycnJYvXo1oaGhHDhwgBEjRjBkyBBMJhNms7nOY65cucLOnTtZunQprVu35vbt2w2e48SJE6Snp9OpUycmTZqEi4sLOTk5ODg4sHLlSqVEbzcAACAASURBVG7cuIG/vz/jx4/n9u3b2Nvbo1KpAHB2dubmzZsP5Vo9PDz4/PPP+f3vf4/BYODs2bN07twZqNplNTo6Go1Gg52dHVFRUTWOzc/P58aNG/j5+dXZd1JSEklJSQDExMQ8lLxCiCeTFMhCCNGEFRcXs2LFCsLCwujSpQs+Pj7s3LmTwsJCBg0aVO/ocVpaGkFBQbRu3RoAR0fHes/Rv39/nnnmGWxsbPj6669JSEhg8eLFmM1m0tPTWbFiBS4uLqxatYpDhw4xYMCAR3KtAAEBAWRlZbFgwQJat26Nj4+PtRD/xz/+wfz58+nRowdffvklf/vb32ps/PTtt98SFBSEUln3/ed6vd66sZQQQjREVrEQQogmzN7ennbt2nH+/HkAnn32WebOnYtGoyEqKoq0tLQ6j7NYLPec5nCXk5MTNjY2QFURmZ2dDVSNDHt5edGhQwdUKhWBgYFkZ2fj5OREWVkZlZWVANy8edN6g199Pv30U8LDwwkPD79nntGjRxMbG8vChQuxWCx07NiR4uJiLl26RI8ePYCqOdN3P5O7jh49yjPPPNOoaxZCiIZIgSyEEE2YWq0mPDycw4cPk5ycTF5eHh06dGDEiBEMGDCAS5cu1Xmcv78/x44do6SkBKDBKRZFRUXWx6dOnbJOaejevTulpaUUFxcDVaPSnTt3RqFQ0KtXL44fPw7QqFHl119/ndjYWGJjYxtsZzabrZkvXbrE5cuXCQgIwMHBgbKyMuvud//5z39wd3e3HpeTk0NpaSk+Pj4N9i+EEI0hUyyEEKKJs7W1JSIigmXLltGvXz+OHTuGSqXiqaeeYuzYsXUe06VLF15++WXee+89lEolnp6ezJw5s862e/bs4dSpU6hUKhwdHQkNDQVAqVQyceJElixZgsVioVu3btYpCuPHj2f16tVs27YNLy8vnnvuOQAyMzNZuXIlpaWlnD59mu3btxMfH1/rnPW1M5lMLFq0CKgaPf/jH/9onWIxbdo04uLiUCqVODg4MGPGDGt/ycnJPP30040eNRdCiIYoLBaL5XGHeFL95YPT92zzu1F1zx98mJrSnulNJUtTyQFNJ0tTyQFNJ8v95HBzc3tEacTDdnc0uilrKt+FhjSHjNA8cjaHjNA8ctaXsbG/0TLFQgghhBBCiGpkioUQQjRjJSUlLFmypNbrixYtwsnJqcZrO3fu5NixYzVeGzx4MKNHj36kGYUQormRAlkIIZoxJyene974dtfo0aOlGBZCiEaQKRZCCCGEEEJUIwWyEEIIIYQQ1cgUCyGEEA+d0Whk8eLFmEwmKisrCQoK4pVXXqnR5uzZsyQmJhIREcH27duxtbVl1KhRNdrcvHmTv/zlL4SFhTV4vvfee4+ioiI0Gg0ACxYsoE2bNg/3ooQQLYYUyEIIIR46GxsbFi9ejK2trXVt4z59+vzijTycnZ3vWRzfNXv2bLy9ve8nrhBC1CAF8iP0a6xxLIQQTZFCocDW1haAyspKKisrUSgUfPfdd2zevBknJye8vLxqHHPp0iUiIyMpLCxk1KhR6PV6bty4wfLly4mLi8NgMJCQkEBOTg7u7u7k5+fz1ltvSVEshHjopEAWQgjxSJjNZubNm0dubi6/+93v8PDwYM6cOSxatIiOHTuyatWqGu0vX75MVFQU5eXlzJs3j379+tV4f9++fTg6OrJy5UouX77M3Llza7y/fv16lEolgwYNYsyYMbV21UtKSiIpKQmAmJiYR3DFQognhRTIQgghHgmlUklsbCylpaWsXLmS7OxsXF1d6dSp6q9rwcHB1oIVYMCAAWg0GjQaDb169SIzMxNPT0/r+xkZGYwYMQKArl274uHhYX1v9uzZODs7c+fOHeLi4jhy5Ai/+c1vauTR6/XWrbKFEKIhsoqFEEKIR8rBwQFfX18uXLjQYLufj/j+/HlDnJ2dAbCzs+PZZ58lMzPzlwcVQoj/nxTIQgghHrri4mJKS0uBqhUtUlNTcXd358aNG+Tm5gKQnJxc45iUlBSMRiMlJSWcPXu21txirVZr3Qnw6tWrXL58Gaia41xcXAyAyWTi9OnTdOnS5ZFenxDiySZTLIQQQjx0RUVFJCQkYDabsVgsDB48mP79+6NSqYiJicHJyQmtVsuVK1esx3Tv3p2YmBgKCgoYM2YMzs7O3Lhxw/r+8OHDSUhI4N1338XT05OuXbtib29PRUUFUVFRVFZWYjab8ff3l6kUQogHorBYLJbHHeJJlZOT87gjAODi4kJBQcHjjgE0nSxNJQc0nSxNJQc0nSz3k8PNze0RpRFmsxmTyYRGoyE3N5elS5eyZs0a1Or7G+tpKr/RDWkq34WGNIeM0DxyNoeM0Dxy1pexsb/RMoIshBCiWTAYDERGRlJZWYnFYmHKlCn3XRwLIURD5JdFCCFEs2BnZyfLswkhfhVyk54QQgghhBDVSIEshBBCCCFENVIgCyGEEEIIUY3MQX6E1q5da308e/bsx5hECNFSRUdHM3v2bBwcHBrV/saNGyxfvpy4uLhHmqukpIT4+HgyMzMZOnQob731FlB1I158fDx5eXkolUr69+/P+PHjATh06BBbtmyxbgoSEhLCb3/7W2ufZWVlvPPOOwQGBlr7E0KI+yEFshBCPMHmz5//uCPUycbGhldffZXLly/XWAsZYOTIkfj5+WEymViyZAn//ve/6du3LwBPP/10vcXvZ599hq+v7yPPLoR48kmBLIQQzdju3buxsbFhxIgRbN68mUuXLrF48WJSU1M5ePAg58+fJzo6mvLycqKjo+nZsycXLlzA2dmZuXPnotFoyM7OZsOGDWg0GrRarbVvo9HIRx99RFZWFiqVijfeeAM/Pz+io6MZN24cHh4ezJ07l8DAQMaOHcu2bdto3749/fr1Y/Xq1ZSVlWE2m5kyZQo6na5GbltbW7RarXVXvbtatWqFn58fAGq1Gi8vLwoLC+/5OWRnZ/PTTz/Rp08fsrKyHsInK4RoyWQOshBCNGM6nY6MjAygqkgsLy/HZDKRkZFRqyi9fv06ISEhxMfHY29vz/HjxwFYv349b775JlFRUTXa79u3D4C4uDjmzJlDQkICRqMRnU5Heno6ZWVlqFQqzp8/D8D58+fR6XQkJycTEBBAbGwssbGxeHp63te1lZaWcvr0afz9/a2vnThxgnfffZe4uDjrJgBms5m//e1vTJgw4b7OI4QQPycFshBCNGPdunUjOzubO3fuYGNjg4+PD9nZ2XUWyK6urtZitVu3buTn51NWVkZpaal1akJwcLC1fUZGhvW5u7s77du35/r162i1WtLT08nIyKBv376Ul5djMBi4ceMGbm5ueHt7c/DgQbZv387ly5exs7P7xddVWVnJmjVreP755+nQoQMA/fv3JyEhgZUrV+Lv709CQgIAX3/9NX379sXFxaXBPpOSkoiIiCAiIuIX5xFCtCwyxUIIIZoxtVpN+/btOXjwID4+Pnh4eJCWlkZubi7u7u412trY2FgfK5VKjEYjFosFhUJRZ98Wi6XO17t37052djYdOnSgd+/elJSUsH//frp16waAr68vkZGRnDlzhnXr1jFq1Cjs7OzYsWMHANOnT8fb27vB69q4cSMdO3bkhRdesL7m5ORkfazX69m6dSsAFy5cID09na+//to6gm5ra2u9ua/6MXq9vsHzCiEESIEshBDNnk6nIzExkRkzZtC1a1f++te/0q1bt3oL3+ocHBywt7cnIyMDrVbLN998Y33P19eXb775Bj8/P3JycigoKMDNzQ21Wk27du04duwYY8aMobi4mC1btjBy5EgA8vPzcXZ2Rq/XYzAYuHjxIpMnTyYwMLBR17Nt2zbKysqYPn16jdeLiopo27YtAKdOnaJz585AzVWCDh06RFZWVq3iWAghfgkpkIUQopnT6XTs2rULHx8fbG1t0Wg0taZXNCQ0NNR6k15AQID19eHDh7Np0ybCwsJQqVSEhoZaR6G1Wi1paWm0atUKrVZLYWGh9Qa/s2fPkpiYiEqlwtbWllmzZtV53pkzZ1JWVobJZCIlJYUFCxZgZ2fHzp07cXd3Z968ecD/Lee2Z88eTp06hUqlwtHRkdDQ0Pv9yIQQokEKS31/QxMPrPo8t8e5DrKLi4v1ZpbHralkaSo5oOlkaSo5oOlkuZ8cbm5ujyiNeNhycnIed4R7airfhYY0h4zQPHI2h4zQPHLWl7Gxv9Fyk54QQgghhBDVSIEshBBCCCFENVIgCyGEEEIIUY0UyEIIIYQQQlQjBbIQQgghhBDVSIEshBBCCCFENbIOshBCtGALFixg2bJlD9THf/7zH7Zu3YrJZEKtVjNx4kT8/PwAyM7OJiEhAaPRSN++fXnzzTdRKBRs2bKF06dPo1ar6dChA6GhoTg4ODTYV3VZWVkcPnyYP/zhDw+UXQgh6iIFshBCtGAPWhxD1RbQ8+bNw9nZmcuXLxMVFcXGjRsB2LRpE9OmTaNHjx5ER0fz3Xff0bdvX3r37s24ceNQqVR88skn7Nq1iwkTJjTYV3Xe3t733K5aCCHulxTIQgjRgk2cOJEtW7ZQVFTE6tWrKSsrw2w2M2XKFHQ6HcnJyezatQuAvn37MmHChFp9eHl5WR936dKFiooKKioquH37Nnfu3MHHxweA4OBgUlJS6Nu3b40d+3x8fDh+/HiDfd3dwe+uu7v1RUREsH37dvLy8rh58yaFhYWMGjUKvV7/8D4kIUSLIwWyEEIIkpOTCQgIYPTo0ZjNZgwGAzdv3mTr1q0sX74cBwcHli1bxsmTJwkMDKy3nxMnTuDl5YWNjQ03b96kXbt21vfatWvHzZs3ax1z4MABnn766Qb7upe7o83l5eXMmzePfv364ezsXKNNUlISSUlJAMTExNyzTyFEyyUFshBCCLy9vdmwYQMmk4nAwEA8PT1JS0ujV69etG7dGoAhQ4aQnp5eb4F85coVtm7dyp///GcALBbLPc+7c+dOVCoVQ4YMabCvexkwYAAajQaNRkOvXr3IzMyslVOv18vIshCiUaRAFkIIga+vL5GRkZw5c4Z169YxatQo7Ozs6mx78uRJduzYAcD06dPx9vamsLCQlStXMnPmTDp27AhUjRgXFhZajyssLKwxqnvo0CFOnz7NokWLUCgUNdr9vK+fn/Pnqh9f13MhhPglpEAWQghBfn4+zs7O6PV6DAYDFy9e5MUXX2Tz5s0UFxfj6OjIt99+S0hICAMHDqwxOltaWkpMTAyvv/46Wq3W+nrbtm2xs7PjwoUL9OjRgyNHjhASEgLAd999x+7du4mMjKRVq1b37CswMLDGOc+ePVsjf0pKCi+99BIGg4GzZ88ybty4h/4ZCSFajgcukKOjo5k9ezYODg6Nan/jxg2WL19OXFzcg566QSUlJcTHx5OZmcnQoUN56623ADAYDMTHx5OXl4dSqaR///6MHz8eqBrN2LJli3WEIyQkhN/+9rf8+OOPbNq0iTt37qBUKhk9enSd8+WEEKK5unvTm0qlwtbWllmzZtG2bVvGjRtHZGQkUHWT3sCBA2sdu3fvXnJzc/niiy/44osvgKrl49q0acOUKVNYv349RqORPn360LdvXwA+/vhjTCYTS5cuBaBHjx5MnTq1wb4a0r17d2JiYigoKGDMmDG15h8LIcQv8cAF8vz58x9GjofOxsaGV199lcuXL3PlypUa740cORI/Pz9MJhNLlizh3//+t/VH++mnn7YW03dpNBpmzZpFp06duHnzJhEREQQEBDT6PwVCCNFUbdmyBYChQ4cydOjQWu8/++yzPPvssw32MWbMGMaMGVPne97e3nUOiKxbt+4X91VdSUkJjo6O1uedOnVi2rRp9zxOCCEa454F8u7du7GxsWHEiBFs3ryZS5cusXjxYlJTUzl48CDnz58nOjqa8vJyoqOj6dmzJxcuXMDZ2Zm5c+ei0WjIzs5mw4YNaDSaGn8yMxqNfPTRR2RlZaFSqXjjjTfw8/MjOjqacePG4eHhwdy5cwkMDGTs2LFs27aN9u3b069fvzqXI6rO1tYWrVZLbm5ujddbtWplXXRerVbj5eVVY45cXdzc3KyPnZ2dadOmDcXFxVIgCyHEY3Dq1Cm2bdvGjBkzHncUIcQT6p4Fsk6n46uvvmLEiBFkZ2dTUVGByWQiIyMDnU7H+fPnrW2vX7/OnDlzmD59OvHx8Rw/fpzg4GDWr1/PH/7wB3x9fa2jFQD79u0DIC4ujmvXrrFs2TLWrFmDTqcjPT2d9u3bo1KprOc4f/48wcHBdS5HdD9KS0s5ffo0I0aMsL524sQJ0tPT6dSpE5MmTcLFxaXGMZmZmZhMJjp06FCrP1lCSAghHr0BAwYwYMAA6/NXXnnlMaYRQjyJlPdq0K1bN7Kzs7lz5w42Njb4+PiQnZ1tLZCrc3V1xdPT03pcfn4+ZWVllJaW4uvrC1QtFH9XRkaG9bm7uzvt27fn+vXraLVa0tPTycjIoG/fvpSXl2MwGLhx4wZubm54e3tz8OBBtm/fzuXLl+u907ohlZWVrFmzhueff95a7Pbv35+EhARWrlyJv78/CQkJNY4pKipi3bp1zJgxA6Wy9ken1+uJiYmR4lgIIYQQohm75wiyWq2mffv2HDx4EB8fHzw8PEhLSyM3Nxd3d/cabasv5q5UKjEajVgslnqX26lvjczu3buTnZ1Nhw4d6N27NyUlJezfv59u3boB9S9H9PNlhxqyceNGOnbsyAsvvGB9zcnJyfpYr9ezdetW6/OysjJiYmJ47bXXrLtCCSGEEEKIJ889R5ChappFYmIiOp0OrVbLv/71Lzw9PRu1zqSDgwP29vZkZGQA8M0331jf8/X1tT7PycmhoKAANzc31Go17dq149ixY/To0aPG+aFqOaI2bdqg1+t57rnnuHjxIoGBgcTGxhIbG3vP4njbtm2UlZUxefLkGq8XFRVZH586dYrOnTsDYDKZWLlyJcHBwQwePPjeH5gQQgghhGi2GrWKhU6nY9euXfj4+GBra4tGo6k1vaIhoaGh1pv0AgICrK8PHz6cTZs2ERYWhkqlIjQ01DoKrdVqSUtLo1WrVmi1WgoLC603+NW1HFFdZs6cSVlZGSaTiZSUFBYsWICdnR07d+7E3d2defPmAf+3nNuePXs4deoUKpUKR0dHQkNDATh69Cjp6emUlJRw6NAha993p5MIIYRonF+y1GdWVhaHDx/mD3/4A2VlZaxdu5bCwkIqKysZOXIkw4YN+xUSCyFaIoWlMXuBivsSERFhfTx79uzHlsPFxYWCgoLHdv7qmkqWppIDmk6WppIDmk6W+8lRfdUbUdv9roW/c+dOysrKmDBhAsXFxcyZM4dNmzahVtc/zlNZWYlKpar3/ZycnF+U4XFoKt+FhjSHjNA8cjaHjNA8ctaXsbG/0bKTnhBCiMciLy+PuLg43nzzTQ4ePFhryc+7fy2MiIhAoVBQXl6OxWKhvLwcR0fHOm+WTkhIwNHRkR9//BEvLy/eeOONx3BlQojmTgpkIYQQv7qcnBxWr15NaGgoqampQO0lP6sLCQlhxYoVTJs2jTt37vDOO+/UWSBD1ZKjCxcurPW+LMUphGgsKZCFEEL8qoqLi1mxYgVhYWF06dKFHTt28PzzzwM1l/ys7vvvv8fDw4NFixaRl5fH0qVL0Wq12Nvb1+o/KCio3qU49Xr9o7koIcQTRQpkIYQQvyp7e3vatWvH+fPn6dKlS71LflZ38OBBXnrpJRQKBR07dsTV1ZWcnBxSUlI4c+YMALGxsUDVTqpCCPEgpEAWQgjxq1Kr1YSHhxMVFYWtra11yU8/P78aS35euHDBeoyLiwupqanodDpu3bpFTk4Orq6uvP7667z++uuP8WqEEE8iKZCFEEL86mxtbYmIiGDZsmWMHj2aS5cu1bnk511jxoxh/fr1hIWFATB+/Hhat279OKILIVoAKZCFEEL8alxdXa1LvDk4OBAdHQ3AwIEDa7Xt1asXvXr1AsDZ2ZkFCxbcs/+ZM2c+xLRCiJaqUTvpCSGEEEII0VJIgSyEEEIIIUQ1UiALIYQQQghRjRTIQgghhBBCVCMFshBCCCGEENXIKhZCCCEAWLBgAcuWLXugPkwmEx988AEXL17EbDYTHBzMyy+/DMBXX33FgQMHUCgUdOnShdDQUDQaDQkJCZw7d866K97MmTPx9PTk9u3bbNiwgby8PGxsbJgxYwZdu3atdc7GthNCiMaSAlkIIQTAAxfHAMePH8dkMhEXF4fBYOBPf/oTzzzzDGq1mj179rBq1So0Gg3x8fEcPXqUoUOHAjBx4kSCgoJq9LVr1y48PT0JDw/n2rVrfPzxxyxatKjWORvbTgghGksKZCGEEEBVkbplyxaKiopYvXo1ZWVlmM1mpkyZgk6nIzk5mV27dgHQt29fJkyYUGc/5eXlVFZWYjQaUavV2NvbYzQaMZvNGI1GVCoVRqORtm3bNpjn6tWr1tFnd3d38vPzuXXrFk899dR9tRNCiMaSAlkIIUQNycnJBAQEMHr0aMxmMwaDgZs3b7J161aWL1+Og4MDy5Yt4+TJkwQGBtY4NigoiFOnTjF16lSMRiOTJk3C0dERgJEjRzJjxgw0Gg0BAQEEBARYj/v000/5/PPP8fPzY/z48djY2ODh4cGJEyfQarVkZmaSn5/PzZs3axW+jW0nhBCNJTfpCSGEqMHb25uDBw+yfft2Ll++jJ2dHVlZWfTq1YvWrVujUqkYMmQI6enptY7NzMxEqVSyceNG3n//fRITE8nLy+P27dukpKSQkJDAxo0bKS8v58iRIwCMGzeO1atXEx0dze3bt9m9ezcAL730EqWlpYSHh7Nnzx68vLxQKmv/s9XYdklJSURERBAREfGQPzEhxJNGRpCFEELU4OvrS2RkJGfOnGHdunWMGjUKOzu7OtuePHmSHTt2ADB9+nSSk5Pp06cParWaNm3a0LNnT7KyslAoFLi6utK6dWsABg0axIULFwgODrZOtbCxsWHYsGEkJiYCYG9vT2hoKAAWi4VZs2bh6urK3r172b9/PwDz58/H2dm5znY/p9fr0ev1D/GTEkI8qaRAFkIIUUN+fj7Ozs7o9XoMBgMXL17kxRdfZPPmzRQXF+Po6Mi3335LSEgIAwcOrDHNIjU1lbS0NIYMGYLBYOCHH37ghRdewGg08sMPP2AwGNBoNKSmpuLt7Q1AUVERbdu2xWKxkJKSQpcuXQAoLS2lVatWqNVq9u/fj06nw97enpCQEEJCQqznrK+dEELcLymQH6HZs2c/7ghCCPGLnT17lsTERFQqFba2tsyaNYu2bdsybtw4IiMjgaqb9AYOHFjr2JCQENavX09YWBgWi4Vhw4bh4eEBVM1PnjdvHiqVCk9PT+to7tq1aykuLgaq5hNPnToVgGvXrvH++++jVCrp3Lkz06dPrzNvY9sJIURjKSwWi+Vxh3hS5eTkPO4IALi4uFBQUPC4YwBNJ0tTyQFNJ0tTyQFNJ8v95HBzc3tEacTD1lR+oxvSVL4LDWkOGaF55GwOGaF55KwvY2N/o+UmPSGEEEIIIaqRAlkIIYQQQohqpEAWQgghhBCiGimQhRBCCCGEqEYKZCGEEEIIIaqRAlkIIYQQQohqZB1kIYQQACxYsIBly5Y9UB8mk4kPPviAixcvYjabCQ4O5uWXXwbgq6++4sCBAygUCrp06UJoaCgajYaEhATOnTtn3dxj5syZeHp6cvv2bTZs2EBeXh42NjbMmDGDrl271jpnSkoKn332GQqFApVKxeTJk9FqtQ90HUKIlk0KZCGEEAAPXBwDHD9+HJPJRFxcHAaDgT/96U8888wzqNVq9uzZw6pVq9BoNMTHx3P06FGGDh0KwMSJEwkKCqrR165du/D09CQ8PJxr167x8ccfs2jRolrn9Pf3Z8CAASgUCi5dusSqVatYvXr1A1+LEKLlkgJZCCEEUFWkbtmyhaKiIlavXk1ZWRlms5kpU6ag0+lITk5m165dQNVOehMmTKizn/LyciorKzEajajVauzt7TEajZjNZoxGIyqVCqPRSNu2bRvMc/XqVevos7u7O/n5+dy6dYunnnqqRjtbW1vrY4PBgEKheJCPQQghpEAWQghRU3JyMgEBAYwePRqz2YzBYODmzZts3bqV5cuX4+DgwLJlyzh58iSBgYE1jg0KCuLUqVNMnToVo9HIpEmTcHR0BGDkyJHMmDEDjUZDQEAAAQEB1uM+/fRTPv/8c/z8/Bg/fjw2NjZ4eHhw4sQJtFotmZmZ5Ofnc/PmzVoFMsDJkyf53//9X3766Sfmz59f53UlJSWRlJQEQExMzMP6uIQQTyApkB8hZXocZl3Y444hhBC/iLe3Nxs2bMBkMhEYGIinpydpaWn06tWL1q1bAzBkyBDS09NrFciZmZkolUo2btxIaWkpixYtwt/fHwcHB1JSUkhISMDe3p74+HiOHDlCcHAw48aN46mnnsJkMrFx40Z2797N2LFjeemll9i8eTPh4eF07doVLy8vlMq67y0PDAwkMDCQc+fO8dlnn7Fw4cJabfR6PXq9/uF/YEKIJ44UyEIIIWrw9fUlMjKSM2fOsG7dOkaNGoWdnV2dbU+ePMmOHTsAmD59OsnJyfTp0we1Wk2bNm3o2bMnWVlZKBQKXF1drQX2oEGDuHDhAsHBwdapFjY2NgwbNozExEQA7O3tCQ0NBcBisTBr1ixcXV3Zu3cv+/fvB2D+/Pk4OzvXyJ6QkEBxcbH1XEII8UtJgSyEEKKG/Px8nJ2d0ev1GAwGLl68yIsvvsjmzZspLi7G0dGRb7/9lpCQEAYOHFhjFDk1NZW0tDSGDBmCwWDghx9+4IUXXsBoNPLDDz9gMBjQaDSkpqbi7e0NQFFREW3btsVisZCSkkKXLl0AKC0tpVWrVqjVavbv349Op8Pe3p6QkBBCQkKss8yiHQAAIABJREFU58zNzaVDhw4oFAqys7MxmUw4OTn9uh+aEOKJIgWyEEKIGs6ePUtiYiIqlQpbW1tmzZpF27ZtGTduHJGRkUDVTXoDBw6sdWxISAjr168nLCwMi8XCsGHD8PDwAKrmJ8+bNw+VSoWnp6d1usPatWspLi4GwMPDg6lTpwJw7do13n//fZRKJZ07d2b69Ol15j1+/DhHjhxBpVKh0Wh455135EY9IcQDUVgsFsvjDvGkyt0f1iTmILu4uFBQUPC4YwBNJ0tTyQFNJ0tTyQFNJ8v95HBzc3tEacTDlpOT87gj3FNT+S40pDlkhOaRszlkhOaRs76Mjf2Nlp30hBBCCCGEqEYKZCGEEEIIIaqRAlkIIYQQQohqpEAWQgghhBCiGimQhRBCCCGEqEaWeRNCiBZmwYIFLFu27IH6MJlMfPDBB1y8eBGz2UxwcDAvv/wyAF999RUHDhxAoVDQpUsXQkND0Wg0JCQkcO7cOezt7QGYOXMmnp6e3L59mw0bNpCXl4eNjQ0zZsyga9eutc557do11q9fz8WLF3nttdcYNWqU9b2ZM2dia2uLUqlEpVLJVtJCiAciBbIQQrQwD1ocQ9XawyaTibi4OAwGA3/605945plnUKvV7Nmzh1WrVqHRaIiPj+fo0aMMHToUgIkTJxIUFFSjr127duHp6Ul4eDjXrl3j448/ZtGiRbXO6ejoyJtvvklKSkqdmRYvXiy75wkhHgopkIUQooWZOHEiW7ZsoaioiNWrV1NWVobZbGbKlCnodDqSk5PZtWsXULUhyIQJE+rsp7y8nMrKSoxGI2q1Gnt7e4xGI2azGaPRiEqlwmg0WreSrs/Vq1eto8/u7u7k5+dz69YtnnrqqRrt2rRpQ5s2bThz5sxD+BSEEKJ+UiALIUQLlZycTEBAAKNHj8ZsNmMwGLh58yZbt25l+fLlODg4sGzZMk6ePFljO2mo2hXv1KlTTJ06FaPRyKRJk3B0dARg5MiRzJgxA41GQ0BAAAEBAdbjPv30Uz7//HP8/PwYP348NjY2eHh4cOLECbRaLZmZmeTn53Pz5s1aBfK9REVFAfBf//Vf1l36qktKSiIpKQlApmAIIRokBbIQQrRQ3t7ebNiwAZPJRGBgIJ6enqSlpdGrVy/rVIUhQ4aQnp5eq0DOzMxEqVSyceNGSktLWbRoEf7+/jg4OJCSkkJCQgL29vbEx8dz5MgRgoODGTduHE899RQmk4mNGzeye/duxo4dy0svvcTmzZsJDw+na9eueHl5oVT+snvIly5dirOzMz/99BPLli3Dzc0NX1/fGm30en2dhbMQQvycFMhCCNFC+fr6EhkZyZkzZ1i3bh2jRo3Czs6uzrYnT55kx44dAEyfPp3k5GT69OmDWq2mTZs29OzZk6ysLBQKBa6urtYCe9CgQVy4cIHg4GDrVAsbGxuGDRtGYmIiAPb29oSGhgJgsViYNWsWrq6u7N27l/379wMwf/58nJ2d672Wu++1adOGgQMHkpmZWatAFkKIxpICWQghWqj8/HycnZ3R6/UYDAYuXrzIiy++yObNmykuLsbR0ZFvv/2WkJAQBg4cWGMUOTU1lbS0NIYMGYLBYOCHH37ghRdewGg08sMPP2AwGNBoNKSmpuLt7Q1AUVERbdu2xWKxkJKSQpcuXQAoLS2lVatWqNVq9u/fj06nw97enpCQEEJCQu55HeXl5VgsFuzs7CgvL+c///kPY8eOfTQfmhCiRZACWQghWqizZ8+SmJiISqXC1taWWbNm0bZtW8aNG0dkZCRQdZPewIEDax0bEhLC+vXrCQsLw2KxMGzYMDw8PICq+cnz5s1DpVLh6elpndawdu1aiouLAfDw8GDq1KlA1fJt77//Pkqlks6dOzN9+vQ68966dYuIiAju3LmDQqHgn//8J/Hx8ZSUlLBy5UoAKisrefbZZ+nTp8/D/bCEEC2KwmKxWB53iCdV7v4wzLqwxx0DFxcXCgoKHncMoOlkaSo5oOlkaSo5oOlkuZ8cbm5ujyiNeNhycnIed4R7airfhYY0h4zQPHI2h4zQPHLWl7Gxv9Gyk54QQgghhBDVSIEshBBCCCFENVIgCyGEEEIIUU2Lvknv0KFD9O7d27o8UHp6Ops2bUKlUhEVFYVGo7G2LSgoICEhgdLSUsxmM+PGjaNfv36PK7oQQgghhHhEWnyB3KVLF2uB/M033zBy5EiGDRtWq+0XX3zB4MGDGT58OFevXiU6OloKZCGEEEKIJ1CLKJBv3LjB8uXLiYuLA+DLL7/k8OHD3Lhxg7Vr16LRaHjuuec4duwY33//PampqcyePbtGHwqFgrKyMgDKysqsC94LIYQQQognS4sokOsyaNAgzp07x8SJE62L2GdnZ9O/f3+CgoJqtf9//+//sWzZMvbu3YvBYGDhwoW12iQlJZGUlARATEzMo70AIYR4RIxGI4sXL8ZkMlFZWUlQUBCvvPJKjTZ311COiIiot5+fD05Ut3btWrKyslCr1Xh7ezN16lTUajUpKSl89tlnKBQKVCoVkydPRqvVAhAVFcUPP/yAVqutcd76+hJCiPslvyCN9O233zJ06FBGjhzJhQsXWLduHXFxcSiV/3efo16vty6IL4QQzZWNjQ2LFy/G1tYWk8nEokWL6NOnDz4+Pg/tHM8++yx//OMfAVizZg0HDhxg+PDh+Pv7M2DAABQKBZcuXWLVqlWsXr0agFGjRmEwGKwDEffqSwgh7leLKJBVKhVms9n6vKKi4p7HfPrpp5w5cwaA2NhYDhw4wH//938D4OPjQ0VFBSUlJbRp0+bRhBZCiMdEoVBga2sLVO1MV1lZiUKh4LvvvmPz5s04OTnh5eVlbV9cXMyaNWu4ffs23t7efPfdd9a/olVWVvL+++/z448/0qlTJ2bNmkWrVq1q3MPRvXt3CgsLAaznBTAYDCgUCutzf39/zp49WytvfX0JIcT9ahHLvLVp04bi4mJKSkqoqKiwFr62trbcuXOnzmNef/11YmNjiY2NBap2ZElLSwPg6tWrVFRU0Lp161/nAoQQ4ldmNpsJDw9nypQp+Pv74+HhwcaNG5k3bx5Llizh1q1b1rY7duzAz8+P5cuXExgYWGP3qpycHPR6PStXrsTOzo59+/bVOI/JZOKbb76psTX0yZMnefvtt4mOjmbGjBmNzlxXX9UlJSURERHR4LQQIYSAFjKCrFarGTNmDP/93/+Nq6urdZvBoUOHsmnTJjQaDVFRUQ328cYbb7Bx40b+8Y9/ABAaGlpjZEMIIZ4kSqWS2NhYSktLWblyJdnZ2bi6utKpUycAgoODrVMdMjIyCA8PB6BPnz44ODhY+2nXrp11DnFwcDD//Oc/GTVqlPX9jz76CJ1Oh06ns74WGBhIYGAg586d47PPPqvzno+61NVXdTINTgjRWC2iQAYYMWIEI0aMqPV69RvyZs6cWe/xnTt3ZunSpY8kmxBCNFUODg74+vpy4cKF+zr+5wMJ1Z/v2LGD4uJipk6dWuexvr6+JCQkUFxcfM+/2N2rLyGE+CVaxBQLIYQQjVdcXExpaSlQtaJFamoq7u7u3Lhxg9zcXACSk5Ot7Xv27MnRo0cB+P77763HQtUmS3eL6+TkZOto8v79+/n+++95++23a9zsnJubi8ViAapWFjKZTDg5OTWYt76+hBDifrWYEWQhhBCNU1RUREJCAmazGYvFwuDBg+nfvz8qlYqYmBicnJzQarVcuXIFqFoGc82aNRw7dgydTkfbtm2xs7OjvLwcd3d3Dh06xIcffkjHjh2tq0ts2rSJ9u3b8+c//xmoWnpz7NixHD9+nCNHjqBSqdBoNLzzzjvWUedFixZx7do1ysvLmT59OtOnT6dPnz719iWEEPdLYbn7X3Xx0OXuD8OsC3vcMXBxcalx08zj1FSyNJUc0HSyNJUc0HSy3E+Ou/c4tCQVFRUolUpUKhUXLlxg06ZN1hucm7KcnJzHHeGemsp3oSHNISM0j5zNISM0j5z1ZWzsb7SMIAshhHggBQUFrFq1CovFglqtZtq0aY87khBCPBApkIUQQjyQTp06sWLFiscdQwghHhq5m0EIIYQQQohqpEAWQgghhBCiGimQhRBCCCGEqEYKZCGEaMESEhI4fvx4rddv3rxJXFzcr5Jh+/btfPnll/ed4b333mPOnDmEh4cTHh7OTz/99ChiCiFaELlJTwghRC3Ozs6EhT3eZSp/SYbZs2fj7e39iBMJIVoKKZCFEKIFOXz4MImJiSgUCrp27YpSqeTcuXN89dVX3Lp1iwkTJhAUFMSNGzdYvnw5cXFxHDp0iFOnTmEwGMjLyyMwMJAJEyYAVTvnbd++HZPJRIcOHQgNDcXW1patW7dy6tQpVCoVvXv35o033qC4uJgPP/yQwsJCACZNmmTdWe/SpUtERkZSWFjIqFGj0Ov1NTIYDAYSEhLIycnB3d2d/Px83nrrLSmKhRCPhBTIQgjRQly5coWdO3eydOlSWrduze3bt/nrX//KrVu3WLJkCTk5OSxfvpygoKBax/7444+sWLECtVrN22+/TUhICBqNhp07d7Jw4UJsbW35+9//zldffUVISAgnT55k9erVKBQK69bTf/nLX/j973+PVquloKCAqKgoVq1aBcDly5eJioqivLycefPm0a9fvxrn37dvH46OjqxcuZLLly8zd+7cGu+vX78epVLJoEGDGDNmjHX3veqSkpJISkoCICYm5qF8pkKIJ5MUyEII0UKkpaURFBRE69atAXB0dARg4MCBKJVKOnfuXO/8XT8/P+zt7QHo3LkzBQUFlJaWcvXqVRYuXAiAyWTCx8cHOzs7NBoNH3zwAf369aN///4ApKamcvXqVWufZWVl3LlzB4ABAwag0WjQaDT06tWLzMxMPD09rW0zMjIYMWIEAF27dsXDw8P63uzZs3F2dubOnTvExcVx5MgRfvOb39S6Br1ej16vv6/PTgjRskiBLIQQLYTFYqlzZNXGxqZGm7pUb6NUKqmsrMRiseDv78/bb79dq/3//M//kJqaytGjR9m7dy+LFy/GYrEQFRWFRqOp1f7nuerKWR9nZ2cA7OzsePbZZ8nMzKyzQBZCiMaSVSyEEKKF8Pf359ixY5SUlABw+/btB+rPx8eH8+fPk5ubC4DBYCAnJ4fy8nLKysro168fkydP5scffwSgd+/e7N2713r83dcBUlJSMBqNlJSUcPbs2Vpzi7VaLceOHQPg6tWrXL58GYDKykqKi4uBqhHs06dP06VLlwe6LiGEkBFkIYRoIbp06cLLL7/Me++9h1KprDGF4X60bt2amTNnsmbNGioqKgB47bXXsLOzY8WKFVRUVGCxWJg0aRIAb775Jh9//DHvvvsulZWV6HQ6pk6dCkD37t2JiYmhoKCAMWPG4OzszI0bN6znGj58OAkJCbz77rt4enrStWtX7O3tqaioICoqisrKSsxmM/7+/jKNQgjxwBSW+v6eJh5Y7v4wzLrHu0wSgIuLCwUFBY87BtB0sjSVHNB0sjSVHNB0stxPDjc3t0eUpmUzm82YTCY0Gg25ubksXbqUNWvWoFbf/zhPTk7OQ0z4aDSV70JDmkNGaB45m0NGaB4568vY2N9oGUEWQgjR5BkMBiIjI61zn6dMmfJAxbEQQjREfl0eoaYweiyEEE8COzs7WZpNCPGrkZv0hBBCCCGEqEYKZCGEEEIIIaqRAlkIIYQQQohqZA6yEEI8waKjo5k9ezYODg6Nan/jxg2WL19OXFzcI81VUlJCfHw8mZmZDB06lLfeeguouhkvPj6evLw8lEol/fv3Z/z48QAcOnSILVu2WDcGCQkJ4be//S0ABQUFfPDBBxQWFgIwf/58XF1dH+k1CCGeXFIgCyHEE2z+/PmPO0KdbGxsePXVV7l8+TJXrlyp8d7IkSPx8/PDZDKxZMkS/v3vf9O3b18Ann76aWsxXd3777/P6NGj6d27N+Xl5b9oJz4hhPg5KZCFEKIZ2717NzY2NowYMYLNmzdz6dIlFi9eTGpqKgcPHuT8+fNER0dTXl5OdHQ0PXv25MKFCzg7OzN37lw0Gg3Z2dls2LABjUaDVqu19m00Gvnoo4/IyspCpVLxxhtv4OfnR3R0NOPGjcPDw4O5c+cSGBjI2LFj2bZtG+3bt6dfv36sXr2asrIyzGYzU6ZMQafT1chta2uLVqu17sJ3V6tWrfDz8wNArVbj5eVlHRWuz9WrV6msrKR3797WvoUQ4kHIHGQhhGjGdDodGRkZAGRnZ1NeXo7JZCIjI6NWUXr9+nVCQkKIj4/H3t6e48ePA7B+/XrefPNNoqKiarTft28fAHFxccyZM4eEhASMRiM6nY709HTKyspQqVScP38egPPnz6PT6UhOTiYgIIDY2FhiY2Pve8e+0tJSTp8+jb+/v/W1EydO8O677xIXF2fdBCAnJwcHBwdWrlzJ3Llz2bJlC2az+b7OKYQQIAWyEEI0a926dSM7O5s7d+5gY2ODj48P2dnZdRbIrq6u1mK1W7du5OfnU1ZWRmlpKb6+vgAEBwdb22dkZFifu7u70759e65fv45WqyU9PZ2MjAz69u1LeXk5BoOBGzdu4Obmhre3NwcPHmT79u1cvnwZOzu7X3xdlZWVrFmzhueff54OHToA0L9/fxISEli5ciX+/v4kJCQAVbvspaenM3HiRKKjo8nLy+PQoUO1+kxKSiIiIoKIiIhfnEcI0bLIFAshhGjG1Go17du35+DBg/j4+ODh4UFaWhq5ubm4u7vXaGtjY2N9rFQqMRqNWCyWeufrWiyWOl/v3r072dnZdOjQgd69e1NSUsL+/fvp1q0bAL6+vkRGRnLmzBnWrVvHqFGjsLOzY8eOHQBMnz4db2/vBq9r48aNdOzYkRdeeMH6mpOTk/WxXq9n69atADg7O+Pl5WUtpAMDA7lw4QLPPfdcjT71ej16vb7B8wohBMgIshBCNHs6nY7ExER0Oh1arZZ//etfeHp6NupGNQcHB+zt7a3TNL755hvre76+vtbnOTk5FBQU4Obmhlqtpl27dhw7dowePXrUOD9Afn4+bdq0Qa/X89xzz3Hx4kUCAwOtUy7uVRxv27aNsrIyJk+eXOP1oqIi6+NTp07RuXNnoKpgLy0tpbi4GIC0tDTre0IIcT9kBFkIIZo5nU7Hrl278PHxwdbWFo1GU2t6RUNCQ0OtN+kFBARYXx8+fDibNm0iLCwMlUpFaGiodRRaq9WSlpZGq1at0Gq1FBYWWm/wO3v2LImJiahUKmxtbZk1a1ad5505cyZlZWWYTCZSUlJYsGABdnZ27Ny5E3d3d+bNmwf833Jue/bs4dSpU6hUKhwdHQkNDQWqRsMnTpzIkiVLsFgsdOvWTUaKhRAPRGGp729o4oHl5OQ87ggAuLi4WG9medyaSpamkgOaTpamkgOaTpb7yeHm5vaI0oiHran8RjekqXwXGtIcMkLzyNkcMkLzyFlfxsb+RssUCyGEEEIIIaqRAlkIIYQQQohqpEAWQgghhBCiGimQhRBCCCGEqEYKZCGEEEIIIaqRAlkIIYQQQohqZB1kIYR4AhmNRhYvXozJZKKyspKgoCBeeeWVGm3urlccERHBqVOnuHr1Ki+99NIDnffHH39k06ZNGI1GVCoVU6ZMoXv37pSVlbF27VoKCwuprKxk5MiRDBs2DIBXX32Vrl27AlVLM91d/zgtLY0tW7ZgMpnw8vJixowZqFSqWufcu3cv//jHP8jLy+Ojjz6idevWD3QNQgghBbIQQjyBbGxsWLx4Mba2tphMJhYtWkSfPn3w8fGps/2AAQMYMGDAA5/3k08+YezYsfTt25czZ87wySef8N5777F37146d+5MREQExcXFzJkzhyFDhqBWq9FoNMTGxtbox2w2k5CQwMKFC3Fzc+Ozzz7j8OHDtbaPBujZsyf9+vUjMjLygfMLIQRIgSyEEE8khUKBra0tAJWVlVRWVqJQKPjuu+/YvHkzTk5OeHl5WdsfOnSIrKws3nrrLY4dO8bnn3+OUqnE3t6eyMhIjEYjH330EVlZWahUKt544w38/PzqPO+dO3cAKCsro23bttbXy8vLsVgslJeX4+joiFJZ/yy/27dvo1arrYv69+7dm7///e91FsjVr0MIIR4GKZCFEOIJZTabmTdvHrm5ufzud7/Dw8ODOXPmsGjRIjp27MiqVavqPO7zzz/nz3/+M87OzpSWlgKwb98+AOLi4rh27RrLli1jzZo1aDSaGsdOmjSJqKgotmzZgtlsZtmyZUDVdtErVqxg2rRp3Llzh3feecdaIFdUVBAREYFKpeLFF18kMDAQJycnKisrycrKwtvbm+PHjz/wzl1JSUkkJSUBEBMT80B9CSGebFIgCyHEE0qpVBIbG0tpaSkrV64kOzsbV1dXOnXqBEBwcLC1YKyuZ8+eJCQkMHjwYAYNGgRARkYGzz//PADu7u60b9+e69ev4+HhUePYr7/+mkmTJhEUFMTRo0f54IMPWLhwId9//z0eHh4sWrSIvLw8li5dilarxd7envXr1+Ps7ExeXh5Lliyha9eudOzYkbfffpu//vWvVFRUEBAQUOf8419Cr9ej1+sfqA8hRMsgq1gIIcQTzsHBAV9fXy5cuNCo9lOnTuW1116jsLCQuXPnUlJSgsViqbPt+vXrCQ8PJzo6GoDDhw9bi+rBgweTmZkJwMGDBxk0aBAKhYKOHTvi6upKTk4OAM7OzgB06NABX19ffvzxRwB8fHxYsmQJ0dHR6HQ6OnbsCEBUVBTh4eF88MEH9/eBCCHEPcgIshBCPIGKi4tRqVQ4ODhgNBpJTU3lxRdfZM+ePeTm5tKxY0eSk5PrPDY3N5cePXrQo0cPTp8+TWFhIb6+vnzzzTf4+fmRk5NDQUEBbm5uhIaG1jjW2dmZc+fO0atXL9LS0qxFrYuLC6mpqeh0Om7dukVOTg6urq7cvn2bVq1aYWNjQ3FxMefPn+fFF18E4KeffqJNmzZUVFSwe/duRo8eDcCf//znR/jJCSGEFMhCCPFEKioqIiEhAbPZjMViYfDgwfTv3x+VSkVMTAxOTk5otVquXLlS69hPPvmE69evA+Dn54eHhwdubm5s2rSJsLAwVCoVoaGh2NjY1Dp22rRp/OUvf8FsNmNjY8O0adMAGDNmDOvXrycsLAyA8ePH07p1a86fP8+HH36IUqnEbDbz0ksv0blzZwC+/PJLzpw5g9lsZvjw4XXeFAjwz3/+ky+//JJbt24RHh5O3759mT59+kP5HIUQLZPCUt/fzcQDu/vnw8fNxcXlgW9ueViaSpamkgOaTpamkgOaTpb7yXF31QXR9DWV3+iGNJXvQkOaQ0ZoHjmbQ0ZoHjnry9jY32iZgyyEEEIIIUQ1UiALIYQQQghRjRTIQgghhBBC/H/t3XlclWX+//EX57AJHBdkEzVQFDi4hIKGNkkLLuNMTsukTmaau2Lqd5TFNFCzREHNDMu0tFG/paktOpP1wNz3JSYVcAHXABFQQRAPZ/n9wc/7CwmCsh308/wnOOc61/2+bx9c96frXPd9lyIFshBCCCGEEKVIgSyEEEIIIUQpcps3IYQQZkGn0xEdHY1er8dgMBAUFMTAgQPLtDl16hRbtmwhMjKynlIKIR4H1S6Q582bx6RJk7C3t69S+6ysLObPn8/ChQuru+n7ys/PZ9GiRZw7d45nn32WkSNHAnDnzh0WLVrE1atXUalUBAQEMGTIEAB27tzJmjVrlKc69evXjxdeeAEoeXLT2bNn8fX1lYFZCCFqgZWVFdHR0dja2qLX64mKisLf3x9vb+/6jiaEeMxUu0CePn16TeSocVZWVgwaNIhLly7dcyP8F198kY4dO6LX65kzZw6//vorXbp0AaBnz55KMV3agAEDuHPnDgkJCXWSXwghHjcWFhbY2toCYDAYMBgMWFhYkJiYyOrVq9FoNLRp00ZpP3XqVObMmYOdnR0jR45k2LBhBAcHs3TpUoKDg+ncuXN97YoQooGrtED+/vvvsbKyon///qxevZqLFy8SHR3NiRMn2LFjB6dPn2bevHkUFRUxb948fHx8OHPmDI6OjoSHh2NtbU1aWhqffPIJ1tbW+Pr6Kn3rdDpWrlxJamoqarWaN998k44dOzJv3jxef/11PDw8CA8Pp3v37vz973/n66+/xtnZma5du/Lhhx9SWFiI0Whk1KhRaLXaMrltbW3x9fUlMzOzzOs2NjbK05gsLS1p06YNOTk5lR6oTp06cerUqSodVCGEEA/HaDQSERFBZmYmffv2xcPDg8mTJxMVFYWbmxuLFy9W2vr4+HD69GmcnJxwdXUlOTmZ4OBgzp49y+jRo+/pOyEhQZnkiImJqbN9EkI0PJVepKfVaklJSQEgLS2NoqIi9Ho9KSkp9xSlGRkZ9OvXj0WLFmFnZ8fBgwcBWLZsGW+99Rbvv/9+mfY//fQTAAsXLmTy5MnEx8ej0+nQarUkJydTWFiIWq3m9OnTAJw+fRqtVsvevXt58skniY2NJTY2Fk9Pz4fa+YKCAo4dO0anTp2U1w4dOsS0adNYuHDhAz8lJiEhgcjISFmCIYQQD0mlUhEbG8unn35KamoqaWlpuLi40KJFCywsLOjVq5fSVqvVkpSURHJyMr179+by5cvk5ubi4OCgzESXFhISQkxMjBTHQohKVVogt23blrS0NG7fvo2VlRXe3t6kpaWVWyC7uLgoxWrbtm25du0ahYWFFBQU4OfnB1BmcEtJSVF+b9myJc7OzmRkZODr60tycjIpKSl06dKFoqIi7ty5Q1ZWFu7u7nh5ebFjxw42bNjApUuXaNSo0QPvuMFgYMmSJfz5z3/G1dUVgICAAOLj44mLi6NTp07Ex8c/UJ8y+AohRM2wt7fHz8+PM2fOVNjm7gROcnIyHTp0QKPRcPDgwTLfVAohxMMMlGQKAAAgAElEQVSotEC2tLTE2dmZHTt24O3tjVar5eTJk2RmZtKyZcsyba2srP6vY5UKg8GAyWTCwsKi3L5NJlO5r7dr104pwv38/PD09GT79u20bdsWAD8/P2bPno2joyNLly5l165dHD58mLCwMMLCwkhNTa10x5cvX46bmxt/+ctflNc0Go2yDyEhIaSlpVXajxBCiJqRl5dHQUEBULIE78SJE7Rs2ZKsrCxludzevXuV9k5OTuTn55OZmYmrqyu+vr5s2bLlnskbIYR4UFW6SE+r1bJlyxbGjx/PE088wZdffknbtm0rLHxLs7e3x87OjpSUFHx9fdmzZ4/ynp+fH3v27KFjx46kp6eTnZ2Nu7s7lpaWNG/enAMHDvDqq6+Sl5fHmjVrePHFFwG4du0ajo6OhISEcOfOHc6fP8/w4cPp3r17lXb666+/prCwkHHjxpV5/fr16zRr1gyAo0eP0qpVqyr1J4QQovquX79OfHw8RqMRk8lEjx49CAgIQK1WExMTg0ajwdfXt8yF1+3atcNoNAIl56qvvvpKZpCFENVW5QL522+/xdvbG1tbW6ytrR/o/9AnTJigXKT35JNPKq/36dOHFStWMHXqVNRqNRMmTFBmcH19fTl58iQ2Njb4+vqSk5OjDHp374OpVquxtbVl4sSJ5W43NDSUwsJC9Ho9R44cYebMmTRq1IjNmzfTsmVLIiIigP+7nduPP/7I0aNHUavVODg4MGHCBKWvqKgofv/9d4qKihg3bhzjxo3D39+/ysdACCHE/Xl4eLBgwYJ7Xvf39+fDDz8s9zNvv/228rOPjw/r16+vtXxCiMeHhamidQ6i2tLT0+s7AlDyNeSDXnBYW8wli7nkAPPJYi45wHyyPEwOd3f3Wkojapq5jNH3Yy5/C/fTEDJCw8jZEDJCw8hZUcaqjtHyqGkhhBBCCCFKkQJZCCGEEEKIUqRAFkIIIYQQohQpkIUQQgghhChFCmQhhBBCCCFKqdJt3oQQQjyadDod0dHR6PV6DAYDQUFBDBw4sEybu7fWjIyMrLS/1NRUdu3axYgRI6qcYcOGDdja2jJgwIAHzi+EELVBCmQhhHiMWVlZER0dja2tLXq9nqioKPz9/fH29n6o/ry8vPDy8qrhlEIIUbekQBZCiMeYhYUFtra2ABgMBgwGAxYWFiQmJrJ69Wo0Gg1t2rRR2k+dOpU5c+ZgZ2fHyJEjGTZsGMHBwSxdupTg4GDUarUy27xhwways7PJysoiOzub/v37079/fwA2b97Mrl27cHJyQqPR0LZtWwAuXLjAihUruHPnDq6urowfPx6DwcAHH3zA/PnzuXDhAuHh4SxbtgwnJyfefvtt4uLiOH78OBs3bkSlUmFnZ8fs2bPr/mAKIR4ZUiALIcRjzmg0EhERQWZmJn379sXDw4PJkycTFRWFm5sbixcvVtr6+Phw+vRpnJyccHV1JTk5meDgYM6ePcvo0aNJTU0t03d6ejrR0dHcvn2bKVOm0KdPHy5dusS+fftYsGABBoOBiIgIpUD++OOPGTFiBH5+fqxfv56NGzcyfPhwiouLKSwsJCUlBS8vL5KTk/H19aVx48bY2NiwceNGZsyYgaOjIwUFBeXuZ0JCAgkJCQDExMTU0tEUQjwKpEAWQojHnEqlIjY2loKCAuLi4khLS8PFxYUWLVoA0KtXL6Ww1Gq1JCUl4ezsTO/evdm+fTu5ubk4ODgoM9Glde3aFSsrK6ysrGjSpAk3b94kOTmZ7t27Y2NjA0BgYCAAhYWFFBQU4OfnB0BwcLBSnHt7e3P69GmSkpJ4+eWXSUxMxGQyodVqgZLCPT4+nh49evDUU0+Vu58hISGEhITU4JETQjyq5C4WQgghALC3t8fPz48zZ85U2Ear1ZKSkkJycjIdOnRAo9Fw8OBBfH19y21vafl/8zAqlQqDwQCULO14EFqtluTkZLKzswkMDOTixYukpKQoBfKYMWMYPHgwOTk5hIeHk5+f/0D9CyFEaVIgCyHEYywvL09ZkqDT6Thx4gQtW7YkKyuLzMxMAPbu3au0d3JyIj8/n8zMTFxdXfH19WXLli1KoVoVWq2Ww4cPo9PpuH37NseOHQPAzs4OBwcHkpOTAdi9e7fSr1arZc+ePbi5uaFSqXBwcODXX3/Fx8cHgMzMTNq3b8+gQYPQaDTk5ORU/+AIIR5bssRCCCEeY9evXyc+Ph6j0YjJZKJHjx4EBASgVquJiYlBo9Hg6+vL5cuXlc+0a9cOo9EIlBSuX331VYUzyOVp27YtPXv2JCwsDGdn5zKfDQ0NVS7Sc3FxYcKECQC4uLgAKMsvfHx8yMnJwcHBAYC1a9eSkZEBQMeOHfHw8KjGURFCPO4sTCaTqb5DPKrS09PrOwJQMuOTnZ1d3zEA88liLjnAfLKYSw4wnywPk8Pd3b2W0oiaZi5j9P2Yy9/C/TSEjNAwcjaEjNAwclaUsapjtCyxEEIIIYQQohQpkIUQQgghhChFCmQhhBBCCCFKkQJZCCGEEEKIUqRAFkIIIYQQohS5zZsQQjRgOp2O6Oho9Ho9BoOBoKAgBg4cWKbNqVOn2LJlC5GRkZX2l5qayq5duxgxYkSVM2zYsAFbW1sGDBjwwPkfxIEDB/jmm2/4/fff+eCDD/Dy8gLgt99+Y926dej1eiwtLRk6dCgdO3as1SxCiEebFMhCCNGAWVlZER0dja2tLXq9nqioKPz9/fH29n6o/ry8vJTC09y0bt2aadOm8dlnn5V5XaPREBERgaOjI5cuXeL9999n+fLl9ZRSCPEokAJZCCEaMAsLC2xtbQEwGAwYDAYsLCxITExk9erVaDQa2rRpo7SfOnUqc+bMwc7OjpEjRzJs2DCCg4NZunQpwcHBqNVqZbZ5w4YNZGdnk5WVRXZ2Nv3796d///4AbN68mV27duHk5IRGo6Ft27YAXLhwQXnQh6urK+PHj8dgMPDBBx8wf/58Lly4QHh4OMuWLcPJyYm3336buLg4jh8/zsaNG1GpVNjZ2TF79ux79rVVq1blHoPS+9e6dWuKi4spLi7Gysqqxo6zEOLxIgWyEEI0cEajkYiICDIzM+nbty8eHh5MnjyZqKgo3NzcWLx4sdLWx8eH06dP4+TkhKurK8nJyQQHB3P27FlGjx5Nampqmb7T09OJjo7m9u3bTJkyhT59+nDp0iX27dvHggULMBgMREREKAXyxx9/zIgRI/Dz82P9+vVs3LiR4cOHU1xcTGFhISkpKXh5eZGcnIyvry+NGzfGxsaGjRs3MmPGDBwdHZVHXz+MQ4cO0aZNm3KL44SEBBISEgCIiYl56G0IIR59UiALIUQDp1KpiI2NpaCggLi4ONLS0nBxcaFFixYA9OrVSykMtVotSUlJODs707t3b7Zv305ubi4ODg7KTHRpXbt2xcrKCisrK5o0acLNmzdJTk6me/fu2NjYABAYGAhAYWEhBQUFyuOgg4ODleLc29ub06dPk5SUxMsvv0xiYiImkwmtVguUFO7x8fH06NGDp5566qGOw+XLl1m3bh0zZswo9/2QkBBCQkIeqm8hxONF7mIhhBCPCHt7e/z8/Dhz5kyFbbRaLSkpKSQnJ9OhQwc0Gg0HDx7E19e33PaWlv83j6JSqTAYDEDJ0o4HodVqSU5OJjs7m8DAQC5evEhKSopSII8ZM4bBgweTk5NDeHg4+fn5LFu2jLCwMObNm1dp/zk5OcTFxREaGoqbm9sDZRNCiD+SAlkIIRqwvLw8ZUmCTqfjxIkTtGzZkqysLDIzMwHYu3ev0t7JyYn8/HwyMzNxdXXF19eXLVu2KIVqVWi1Wg4fPoxOp+P27dscO3YMADs7OxwcHEhOTgZg9+7dSr9arZY9e/bg5uaGSqXCwcGBX3/9FR8fHwAyMzNp3749gwYNQqPRkJOTw4QJE4iNjWX69On3zVNQUEBMTAz/+Mc/Kiz0hRDiQcgSCyGEaMCuX79OfHw8RqMRk8lEjx49CAgIQK1WExMTg0ajwdfXl8uXLyufadeuHUajESgpXL/66qsHKizbtm1Lz549CQsLw9nZucxnQ0NDlYv0XFxcmDBhAgAuLi4AyvILHx8fcnJycHBwAGDt2rVkZGQA0LFjRzw8PO7Z7uHDh/niiy/Iy8sjJiYGT09PZsyYwbZt28jMzGTTpk1s2rQJgJkzZ9KkSZMq75MQQpRmYTKZTPUd4lGVnp5e3xGAkhmj7Ozs+o4BmE8Wc8kB5pPFXHKA+WR5mBzu7u61lEbUNHMZo+/HXP4W7qchZISGkbMhZISGkbOijFUdo2WJhRBCCCGEEKVIgSyEEEIIIUQpUiALIYQQQghRihTIQgghhBBClCIFshBCCCGEEKXIbd6EEOIxptPpiI6ORq/XYzAYCAoKYuDAgWXanDp1ii1bthAZGcnRo0e5cuUKL730UrW2e+HCBVasWIFOp0OtVjNq1CjatWtHYWEhH330ETk5ORgMBl588UWee+45AAYNGsQTTzwBlFyhHhERAcDJkydZs2YNer2eNm3aMH78eNRqdbXyCSEeb1IgCyHEY8zKyoro6GhsbW3R6/VERUXh7++Pt7d3ue0DAwOVR0tXx9q1a/n73/9Oly5dOH78OGvXrmXWrFls27aNVq1aERkZSV5eHpMnT+aZZ57B0tISa2trYmNjy/RjNBqJj4/n3Xffxd3dnfXr17Nr1y6ef/75amcUQjy+pEAWQojHmIWFBba2tgAYDAYMBgMWFhYkJiayevVqNBoNbdq0Udrv3LmT1NRURo4cyYEDB9i4cSMqlQo7Oztmz56NTqdj5cqVpKamolarefPNN+nYsWO52719+zYAhYWFNGvWTHm9qKgIk8lEUVERDg4OqFQVrwa8desWlpaWyr1NO3fuzHfffScFshCiWqRAFkKIx5zRaCQiIoLMzEz69u2Lh4cHkydPJioqCjc3NxYvXlzu5zZu3MiMGTNwdHRUHnf9008/AbBw4UJ+//135s6dy5IlS7C2ti7z2WHDhvH++++zZs0ajEYjc+fOBaBfv34sWLCAsWPHcvv2bf7nf/5HKZCLi4uJjIxErVbzt7/9je7du6PRaDAYDKSmpuLl5cXBgwcrfIBBQkICCQkJAMTExFT/wAkhHllSIAshxGNOpVIRGxtLQUEBcXFxpKWl4eLiQosWLQDo1auXUliW5uPjQ3x8PD169OCpp54CICUlhT//+c8AtGzZEmdnZzIyMu55dPTPP//MsGHDCAoKYv/+/Xz66ae8++67/Pe//8XDw4OoqCiuXr3Ke++9h6+vL3Z2dixbtgxHR0euXr3KnDlzeOKJJ3Bzc2PKlCl8+eWXFBcX8+STT1a4/jgkJISQkJCaPHRCiEeU3MVCCCEEAPb29vj5+XHmzJkqtR8zZgyDBw8mJyeH8PBw8vPzMZlM5bZdtmwZYWFhzJs3D4Bdu3YpRXWPHj04d+4cADt27OCpp57CwsICNzc3XFxclEdCOzo6AuDq6oqfnx8XLlwAwNvbmzlz5jBv3jy0Wi1ubm4PfQyEEAKkQBZCiMdaXl6esjxCp9Nx4sQJWrZsSVZWFpmZmQDs3bu33M9mZmbSvn17Bg0ahEajIScnBz8/P/bs2QNAeno62dnZuLu7M2HCBGJjY5k+fTpQUuwmJSUBJXehuFvUOjk5ceLECQBu3LhBeno6Li4u3Lp1i+LiYiXz6dOnadWqFQA3b94ESpZgfP/99/Tp06fGj5MQ4vEiSyyEEOIxdv36deLj4zEajZhMJnr06EFAQABqtZqYmBg0Gg2+vr5cvnz5ns+uXbuWjIwMADp27IiHhwfu7u6sWLGCqVOnolarmTBhAlZWVvd8duzYsaxatQqj0YiVlRVjx44F4NVXX2XZsmVMnToVgCFDhtC4cWNOnz7NZ599hkqlwmg08tJLLykF8g8//MDx48cxGo306dOn3IsChRDiQViYKvo+TFTb3a8F65uTk1OFF63UNXPJYi45wHyymEsOMJ8sD5Pj7t0UhPkzlzH6fszlb+F+GkJGaBg5G0JGaBg5K8pY1TFallgIIYQQQghRihTIQgghhBBClCIFshBCCCGEEKVIgSyEEEIIIUQpUiALIYQQQghRihTIQgghhBBClCIFshBCPMZ27tzJ559/Xt8xyti6dSv//Oc/mTp1Kh9++CE6nQ6A0NBQ8vLy7mn/888/s2vXLqBkf3Jzc+s0rxDi0SMFshBCCLORm5vLjz/+SExMDAsXLsRoNLJ///77fqZPnz4EBwcDJQXy9evX6yKqEOIRJk/SE0KIR0hWVhbz589n4cKFQMlT5oqKikhKSqJdu3acOnWKwsJCxo0bh1arLfPZ48ePs2nTJiIiIlizZg2NGjUiLS2NGzdu8MYbbxAUFITJZGLt2rUkJiYCJU++69mzJytXrsTf35/AwEBiY2Oxt7dnwoQJ/PLLL2RlZfH8888zb948fHx8OHPmDI6OjoSHh2NtbX3PPhiNRnQ6HWq1Gp1OR7NmzZT3fvjhB06dOgXA5MmTcXNzY8OGDdja2uLi4kJqaiofffQR1tbWvP/+++X2L4QQlZEZZCGEeEwYjUbmzZvHsGHD2LhxY5n3Dh8+zHfffcf06dNp3LgxADdu3GDOnDlERkaybt06AA4dOsSFCxeIjY3l3XffZc2aNVy/fh2tVktycjJQMgv8+++/A5CSkoKvry8AGRkZ9OvXj0WLFmFnZ8fBgwfvyejo6MiLL77I+PHjGTNmDHZ2djz55JPK+3Z2dsybN49+/fqxevXqMp8NCgrCy8uLSZMmERsbe09xnJCQQGRkJJGRkdU4ikKIx4EUyEII8Zjo3r07AG3btiUrK0t5/dSpU3z//fdERkbi4OCgvN6tWzdUKhWtWrXi5s2bQEnB+/TTT6NSqWjatCl+fn6kpqai1WpJSUnhypUrtGrViiZNmnD9+nXOnDmDj48PAC4uLnh6eioZrl27dk/GW7duceTIEeLj41m+fDlFRUXs3r1bef/pp59W/nvmzJkH2v+QkBBiYmKIiYl5oM8JIR4/dbLEYubMmcydO7daffz222+sW7cOvV6PpaUlQ4cOpWPHjgCkpaURHx+PTqejS5cuvPXWW1hYWLBmzRqOHTuGpaUlrq6uTJgwAXt7+/v2VVpqaiq7du1ixIgR1couhBB1Ra1WYzQald+Li4uVn62srABQqVRl2ri4uJCVlUVGRgZeXl73tAcwmUz33a6joyO3bt0iMTERrVbLrVu3OHDgALa2tjRq1Ij8/Pwy/alUKnQ6HdnZ2cyfPx+A3r17o9FocHFxUWaxn3rqKc6cOUOvXr0AsLCwUPoo/bMQQtSkOplBrm5xDKDRaIiIiGDhwoWEhoaydOlS5b0VK1YwduxYPvroIzIzM5W1cZ07d2bhwoXExcXRokULvv3220r7Ks3Ly0uKYyFEg9KkSRPy8vLIz8+nuLiY48ePV/oZZ2dnpk2bxscff8zly5fv21ar1XLgwAGMRiN5eXkkJyfTrl07ALy9vfn3v/+Nn58fWq2WLVu2KMsrKuLk5ERsbCyxsbH06dMHJycnzp49y507dzCZTJw4cYKWLVsq7e9esLd//37at29/T3+2trbcvn270n0WQoj7qZMZ5KFDhyrr1D788EMKCwsxGo2MGjUKrVbL3r17leK1S5cuvPHGG/f00aZNG+Xn1q1bU1xcTHFxMbdu3eL27dt4e3sD0KtXL44cOUKXLl3KrFvz9vZW1rtV1Ffp2Q0o+dpxy5YtREZGsmHDBq5evUpubi45OTkMGDCAkJCQmjtIQghRAywtLXn11Vd55513cHFxwd3dvUqfc3d3Z9KkSSxatIiIiIgK23Xv3p0zZ84QFhYGwBtvvEHTpk2BkuL5t99+w83NDScnJ27dunXPhYCVad++PUFBQURERKBWq/H09Cwz1hYXF/POO+9gMpmYPHnyPZ9/9tlnWbFihVykJ4Soljq9i8XevXt58skneeWVVzAajdy5c4fc3FzWrVvH/Pnzsbe3Z+7cuRw+fFhZK1eeQ4cO0aZNG6ysrMjNzaV58+bKe82bNy/3Hpi//PILPXv2vG9flbl06RLvv/8+RUVFRERE0LVrVxwdHZX3ExISSEhIAJA1bkKIetO/f3/69+9f4fuNGzcmPj4eKCkon332WaBk8mDx4sVAyT2HS1uzZg1Qsqxh6NChDB069J5+n3/+eZ5//nmgpFC/+xkoWcZx984aAAMGDKgw38CBAxk4cOA9r9/N/Nprr93T/q6goCCCgoIq7FsIIaqiTgtkLy8vPvnkE/R6Pd27d8fT05OTJ0/SoUMHZb3ZM888Q3JycoUF8uXLl1m3bh0zZswAKl8XB7B582bUajXPPPPMffuqTGBgINbW1lhbW9OhQwfOnTtXJmdISIjMKgshhBBCNHB1WiD7+fkxe/Zsjh8/ztKlSxkwYACNGjUqt+3hw4f55ptvABg3bhxeXl7k5OQQFxdHaGgobm5uQMmMcU5OjvK5nJycMrO6O3fu5NixY0RFRZW5oKO8vv64zT/64wUhcoGIEEIIIcSjp04L5GvXruHo6EhISAh37tzh/Pnz/O1vf2P16tXk5eXh4ODAvn376NevH926dSszO1tQUEBMTAz/+Mc/ylz00axZMxo1asSZM2do3749u3fvpl+/fgAkJiby/fffM3v2bGxsbCrtq3v37mW2efdm9HcdOXKEl156iTt37nDq1Clef/31Gj9GQgghhBCiftVpgXz3oje1Wo2trS0TJ06kWbNmvP7668yePRsouUivW7du93x227ZtZGZmsmnTJjZt2gSU3D6uSZMmjBo1imXLlqHT6fD396dLly4AfP755+j1et577z2g5OKPMWPG3Lev+2nXrh0xMTFkZ2fz6quvlpmpFkIIIYQQjwYLU1UW8T6mDh48yNGjR5k4caLyKNP7XVjyR+np6bWYruqcnJzIzs6u7xiA+WQxlxxgPlnMJQeYT5aHyVHVu0aI+mcuY/T9mMvfwv00hIzQMHI2hIzQMHJWlLGqY7Q8Sa8CR48e5euvv6Z37971HUUIIYQQQtShOl1i0ZAEBgYSGBio/F7eLYeEEKI+zJo1i6FDh5Z56h3Anj17+P7774GSB2aMGjUKT09PdDod0dHR6PV6DAYDQUFBypi2bds2/v3vf3P16lVWrlyp3FGoor4eRHx8PAEBAXV227XDhw/j7u5Oq1at6mR7QohHlxTIQgjxiHBxcWHWrFk4ODjw66+/8tlnn/HBBx9gZWVFdHQ0tra26PV6oqKi8Pf3x9vbGx8fH7p27apcB1JZX3XFZDJhMplQqar+ReeRI0cICAiQAlkIUW1SIAshhBnbuHEje/fupXnz5mg0Gtq2bQuUzPCuWrWK27dvM378eNq1a4ePj4/yufbt2yu3wLSwsMDW1hYAg8GAwWBQblNZ+smipVXUV0V27drFli1bsLCw4IknnuDtt98GICkpia1bt3Ljxg3eeOMNgoKCKCoqYsGCBRQUFKDX6xk8eDDdunUjKyuLefPm0aFDB+VpfRs2bCAtLQ2A5557jr/+9a9kZmby+eefk5eXh42NDWPHjuXWrVscPXqUpKQkNm3axNSpU5VbeN4lD3MSQlSVFMhCCGGmUlNTOXToEAsWLMBgMBAREaEUyEVFRcydO5ekpCQ++eSTMk+pg5Knh969ow+A0WgkIiKCzMxM+vbtS/v27auc4499/dHly5fZvHkz7733Ho0bN+bWrVvKezdu3GDOnDmkp6czf/58goKCsLKyYtq0adjZ2ZGXl8eMGTOUJW3p6emMHz+eUaNGkZaWRm5urrJvBQUFAHz22WeMHj2aFi1acPbsWVauXEl0dDSBgYH3XdIhD3MSQlSVFMhCCGGmUlJS6NatG9bW1gAEBAQo7/3pT38CSh7AVFhYSEFBAfb29gCcPHmSHTt2MGfOHKW9SqUiNjaWgoIC4uLiuHTpEk888USlGcrrq7w2QUFByvplBwcH5b1u3bqhUqlo1aoVN2/eBEqWT3z11VckJydjYWFBbm6u8p6TkxPe3t5AyTKPrKwsvvjiC7p27Urnzp0pKiri9OnTLFq0SNmGXq+vdD+EEOJBSIEshBBm6kHuwnl3ycTFixdZvnw506dPR6PR3NPO3t4ePz8/EhMTKy2QK+urdM6KnixqZWVVph3A3r17ycvLIyYmBktLS0JDQ9HpdADKUhAoKbRjY2NJTExk27Zt7N+/n+HDh2Nvb09sbOx9swshRHXIbd6EEMJM+fr6cuzYMXQ6HUVFRRw/flx5b//+/UDJLLOdnR12dnZkZ2cTFxfHxIkTy9zrMy8vT1meoNPpOHHiBC1btrzvtivqqzydOnXiwIED5OfnA5RZYlGewsJCmjRpgqWlJSdPnuTatWvltsvLy8NoNBIUFMTgwYM5f/48dnZ2uLi4cODAAaCk6L5w4QIAjRo14vbt2/fdthBCVIXMIAshhJlq164dAQEBhIWF4ezsjJeXF3Z2dkDJ7OrMmTOVi/Sg5IK+W7dusXLlSgDUajUxMTFcv36d+Ph4jEYjJpOJHj16KMs1/vOf//DDDz9w48YNwsLC6NKlC+PGjauwr/K0bt2al19+mVmzZqFSqfD09CQ0NLTC/frTn/7E/PnziYyMxNPTs8JiPTc3l08++QSj0QjA66+/DsCkSZNYsWIFmzdvRq/X8/TTT+Pp6UnPnj1Zvnw5P/74I//85z/vuUhPCCGqSp6kV4vM5SlN5vTEG3PJYi45wHyymEsOMJ8s5vAkvaKiImxtbblz5w7R0dGMGTNGuVBPVI+5jNH3Yy5/C/fTEDJCw8jZEDJCw8hZ3SfpyQyyEEKYseXLl3PlyhWKi4sJDg6W4lgIIeqAFMhCCGHGJk+eXN8RFPn5+eXezSIqKuq+F/EJIURDIwWyEEKIKtFoNHL3CCHEY0HuYiGEEEIIIUQpUiALIYRoUC7Wz0EAAA7nSURBVDZs2MAPP/xQ3zGEEI8wKZCFEEI8cgwGQ31HEEI0YLIGWQghRK3avXs3P/74I3q9nvbt2zNq1CiGDRvGmjVrADh48CDHjh0jNDSUGzdusGLFCrKysgAYNWoUPj4+FfadkJDA4cOHmTZtGh988AHe3t6cPn2awMBAXnzxxTrZPyHEo0cKZCGEELXmypUr7N+/n/feew9LS0tWrlzJnj17Kmy/atUq/Pz8CAsLw2g0UlRUVGHbbdu28d///pewsDDlkdaFhYXMnj27xvdDCPF4kQJZCCFErTl58iTnz59n+vTpQMmjrhs3bnzf9hMnTgRApVIpTw78oz179uDo6EhYWBiWlv93KuvZs2eFfSckJJCQkABQ4VMBhRACpEAWQghRi0wmE8HBwcpjou/aunWr8rNOp7tvH9u2bWP79u0ASqHdunVrLly4QG5uLi4uLkpbGxubCvsJCQkhJCTkgfdBCPH4kYv0hBBC1JpOnTpx8OBBbt68CcCtW7e4du0aTZo04cqVKxiNRg4fPlym/c8//wyA0WiksLCQfv36ERsbS2xsLI6OjgB4enoyZswY5s+fT25ubt3vmBDikSYzyEIIIWpNq1atGDx4MHPnzsVkMqFWqxk5ciRDhgxh/vz5NG/enNatWytrjYcPH85nn33GL7/8gkqlYvTo0Xh7e5fbt6+vL0OHDiUmJoaZM2fW5W4JIR5xUiALIYSoVT179ix3bXBQUNA9rzVt2pTw8PD79jdw4EDlZ39/f/z9/QGYNWtW9YIKIcT/J0sshBBCCCGEKEUKZCGEEEIIIUqxMJlMpvoOIYQQQgghhLmQGeRaEhkZWd8RFJLlXuaSA8wni7nkAPPJYi45RM1rKP+2DSFnQ8gIDSNnQ8gIDSNndTNKgSyEEEIIIUQpUiALIYQQQghRinqW3Ben1rRt27a+Iygky73MJQeYTxZzyQHmk8Vccoia11D+bRtCzoaQERpGzoaQERpGzupklIv0hBBCCCGEKEWWWAghhBBCCFGKFMhCCCGEEEKUIo+arqbExERWrVqF0WjkhRde4KWXXirzfnFxMR9//DFpaWloNBqmTJmCi4tLvWRJSkriyy+/5OLFi0yZMqXcx7zWRY6tW7eyfft21Go1jRs3Zvz48Tg7O9dLlp9//pmffvoJlUqFra0tY8eOpVWrVnWe466DBw+yaNEi5s2bh5eXV43nqEqWnTt3smbNGhwdHQHo168fL7zwQp3nANi/fz/ffPMNFhYWeHh4MHny5BrPUZUsq1ev5tSpUwDodDpu3rzJ6tWrayWLqFnmNEY/bMa6Grurm7Mux/aHzVhXY351c95VF+eEh81YV+eK6uaEhzyXmMRDMxgMpokTJ5oyMzNNxcXFpmnTppkuX75cps22bdtMy5cvN5lMJtPevXtNixYtqrcsV69eNV24cMG0dOlS04EDB+otx4kTJ0xFRUUmk8lk+umnn+r1mBQUFCg/HzlyxDR37tx6yWEymUyFhYWmqKgo0zvvvGM6d+5cjeeoapYdO3aYVq5cWSvbf5Ac6enpprCwMFN+fr7JZDKZbty4UW9ZSvvPf/5jio+Pr5UsomaZ0xhdnYx1MXbXRM66Gturk7EuxvzKmNM5oToZ6+JcUZnaPJfIEotqOHfuHG5ubri6umJpaUnPnj05cuRImTZHjx7l2WefBSAoKIiTJ09iqoXrIquSxcXFBQ8PDywsLGp8+w+So2PHjtjY2ADQvn17cnNz6y2LnZ2d8nNRUVGtHJuq5ABYv349AwYMwMrKqsYzPGiW2laVHNu3b6dv3744ODgA0KRJk3rLUtq+ffv405/+VCtZRM0ypzG6OhnrYuyujDmN7dXJWBdjfmXM6ZxQ3Yz1rTbPJVIgV0Nubi7NmzdXfm/evPk9A0LpNmq1Gjs7O/Lz8+slS1140By//PIL/v7+9Zpl27ZtvP3226xbt4633nqrXnKcP3+e7OxsAgICanz7D5oF4NChQ0ybNo2FCxeSnZ1dLznS09PJyMjg3XffZcaMGSQmJtZ4jqpmuevatWtkZWXRsWPHWskiapY5jdHVyWgOzGlsr4i5jPmVMadzQkXM5VxRmdo8l0iBXA3lzTL88f9Gq9KmrrLUhQfJsXv3btLS0hgwYEC9ZunXrx9Lly5lyJAhbNq0qc5zGI1GvvzyS958880a3/aDZgEICAggPj6euLg4OnXqRHx8fL3kMBqNZGRkEB0dzeTJk/n0008pKCiolyx37du3j6CgIFQqGTobAnMaoytS39uvKnMa2ytiLmN+ZczpnFARczlXVKY2zyUyyldD8+bNycnJUX7PycmhWbNmFbYxGAwUFhYq0/x1naUuVDXHb7/9xrfffkt4eHitfX30oMektr5CqixHUVERly9fZvbs2YSGhnL27FkWLFhAampqnWcB0Gg0yr9JSEgIaWlp9ZLD0dGRbt26YWlpiYuLC+7u7mRkZNRLlrv279/P008/XeMZRO0wpzG6OhnNgTmN7RUxlzG/MuZ0TnjYjFA354rK1Oa5RArkavDy8iIjI4OsrCz0ej379+8nMDCwTJuAgAB27twJlFyN2qFDh1qZHahKlrpQlRznz59nxYoVhIeH19q60qpmKf1Hcvz4cVq0aFHnOezs7Pj888+Jj48nPj6e9u3bEx4eXitXLFflmFy/fl35+ejRo7VyhXdVcnTv3p2TJ08CkJeXR0ZGBq6urvWSBUq+pisoKMDb27vGM4jaYU5jdHUymgNzGturk7EuxvzKmNM54WEzQt2cKypTm+cSeZJeNR0/fpwvv/wSo9HIc889xyuvvML69evx8vIiMDAQnU7Hxx9/zPnz53FwcGDKlCm1cpKvSpZz584RFxdHQUEBVlZWNG3alEWLFtV5jvfee49Lly7RtGlTAJycnIiIiKjxHFXJsmrVKk6cOIFarcbBwYERI0bQunXrOs9R2qxZsxg6dGitDYaVZfnf//1fjh49qhyTUaNG0bJlyzrPYTKZ+Ne//kViYiIqlYpXXnml1mZvq/Lvs2HDBoqLixkyZEitZBC1w5zG6IfNWFdjd3Vz1uXY/rAZ62rMr27O0mr7nPCwGevqXFHdnA97LpECWQghhBBCiFJkiYUQQgghhBClSIEshBBCCCFEKVIgCyGEEEIIUYoUyEIIIYQQQpQiBbIQQgghhBClSIEshJkbOnQoV69ere8YQgghyiFj9KNJbvMmHgmhoaHcuHGjzCOAlyxZgqOjYz2mEkIIATJGi4bHsr4DCFFTIiIi6Ny5c4XvGwwG1Gp1HSYSQghxl4zRoiGRAlk8srKyspg4cSLjxo3jm2++wcXFhdmzZ3PmzBn+9a9/ceXKFZydnRk+fDgdOnRQPhMfH8/58+dp37497u7uFBQUMGnSJE6dOsXSpUv59NNPlW2EhoYyduxYOnfujNFo5IcffmD79u0UFBTQsWNHxowZg4ODg5JlwoQJrF+/Hp1Ox1/+8hdeeeUVAIxGI9999x07duzg5s2btGjRgrCwMJycnBg4cCAfffQRbm5uFBcX89VXX3HgwAH0ej3dunVj+PDhWFtbk5eXx7Jly0hJScHCwoLWrVsza9asMjM2QghhLmSMljHanEmBLB55SUlJLF68GJVKRW5uLjExMUycOBF/f39OnjzJwoUL+fDDD2ncuDFLlizB29ubmTNncvbsWWJiYu557GdFfvzxR44cOcKsWbNo3Lgxq1atYuXKlUyZMkVpk5KSwpIlS0hPT+edd96he/futGrViq1bt7Jv3z6mT59OixYtuHjxIjY2NvdsY926dVy9epXY2FjUajVLlixh48aNvP7662zduhVHR0dWrlwJwNmzZ7GwsKiZgyiEELVExmhhjuR/W8QjIzY2luHDhzN8+HAWLFigvP7aa69ha2uLtbU1u3fvpkuXLnTt2hWVSkXnzp3x8vLi+PHjZGdnk5qayqBBg7CyssLPz4+AgIAqbz8hIYHBgwfTvHlzrKyseO211zh06BAGg6FMFmtrazw9PfHw8ODixYsAbN++ncGDB+Pu7o6FhQWenp5oNJoy/ZtMJrZv386wYcNwcHCgUaNGvPLKK+zbtw8AtVrNjRs3yM7OxtLSEq1WK4OvEMJsyBgtY3RDIjPI4pERFhZWZn1bVlYWAM2bN1dey87O5uDBgxw7dkx5zWAw0KFDB3Jzc7G3t8fW1lZ5z9nZmezs7Cpt/9q1a8TFxZUZ8FQqFTdv3lR+b9q0qfKzjY0NRUVFAOTk5ODq6nrf/vPy8rhz5w6RkZHKayaTCaPRCMCAAQP45ptvmDt3LgAhISG89NJLVcouhBC1TcZoGaMbEimQxSOv9GDYvHlznnnmGcaNG3dPu2vXrlFQUEBRUZEyAJceeG1sbLhz547yu9FoJC8vr0zf48ePx9fX956+754IKtK8eXOuXr3KE088UWEbjUaDtbU1ixYtKvfK70aNGvHmm2/y5ptvcvnyZWbPno2XlxedOnW677aFEKI+yRgtY7Q5kiUW4rHyzDPPcOzYMRITEzEajeh0Ok6dOkVOTg7Ozs54eXmxYcMG9Ho9KSkpZWYx3N3dKS4u5vjx4+j1ejZt2kRxcbHyfu/evfn666+5du0aUDKbcOTIkSrleuGFF1i/fj0ZGRmYTCYuXrxIfn5+mTYqlYoXXniB1atXKzMeubm5JCYmAnDs2DEyMzMxmUw0atQIlUolF38IIRoUGaOFuZAZZPFYcXJyIjw8nLVr17JkyRJUKhXt2rVj9OjRAEyaNIn4+HjeeustvL296dWrFwUFBQDY2dkxatQoPv30U4xGIwMGDCjz1WD//v0BmDt3LtevX6dJkyb06NGDbt26VZrrr3/9K8XFxcydO5f8/HxatmzJtGnT7mk3ZMgQNm7cyIwZM8jPz8fR0ZHevXvj7+9PRkYGX3zxBXl5edjb29OnTx/lym8hhGgIZIwW5kIeFCLEfWzYsIHMzEwmTZpU31GEEEL8gYzRorbI3L4QQgghhBClSIEshBBCCCFEKbLEQgghhBBCiFJkBlkIIYQQQohSpEAWQgghhBCiFCmQhRBCCCGEKEUKZCGEEEIIIUqRAlkIIYQQQohS/h9KvOl8ggI7sgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x504 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "value_freq_graph(X0, X1, figsize=(10, 7))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Preprocessing the data in prevision of ML**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Definitions of a few functions and a class to preprocess the raw data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_header_subject_message(email):\n",
    "    pattern = '\\n{2}'\n",
    "    separator = re.split(pattern, email, maxsplit=1)\n",
    "    header, message = separator[0], separator[1]\n",
    "    subjects = re.findall('Subject:\\s(.*)\\n', header)\n",
    "    if len(subjects) != 0:\n",
    "        subject = subjects[0]\n",
    "    else:\n",
    "        subject = ''\n",
    "    return header, subject, message"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Re: [zzzzteana] Which Muppet Are You?\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "From timc@2ubh.com  Thu Aug 22 17:31:00 2002\n",
      "Return-Path: <timc@2ubh.com>\n",
      "Delivered-To: zzzz@localhost.netnoteinc.com\n",
      "Received: from localhost (localhost [127.0.0.1])\n",
      "\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id A97BE43F99\n",
      "\tfor <zzzz@localhost>; Thu, 22 Aug 2002 12:30:58 -0400 (EDT)\n",
      "Received: from phobos [127.0.0.1]\n",
      "\tby localhost with IMAP (fetchmail-5.9.0)\n",
      "\tfor zzzz@localhost (single-drop); Thu, 22 Aug 2002 17:30:58 +0100 (IST)\n",
      "Received: from n17.grp.scd.yahoo.com (n17.grp.scd.yahoo.com\n",
      "    [66.218.66.72]) by dogma.slashnull.org (8.11.6/8.11.6) with SMTP id\n",
      "    g7MGOFZ14344 for <zzzz@spamassassin.taint.org>; Thu, 22 Aug 2002 17:24:16 +0100\n",
      "X-Egroups-Return: sentto-2242572-52744-1030033454-zzzz=spamassassin.taint.org@returns.groups.yahoo.com\n",
      "Received: from [66.218.67.201] by n17.grp.scd.yahoo.com with NNFMP;\n",
      "    22 Aug 2002 16:24:16 -0000\n",
      "X-Sender: timc@2ubh.com\n",
      "X-Apparently-To: zzzzteana@yahoogroups.com\n",
      "Received: (EGP: mail-8_1_0_1); 22 Aug 2002 16:24:14 -0000\n",
      "Received: (qmail 12796 invoked from network); 22 Aug 2002 16:24:13 -0000\n",
      "Received: from unknown (66.218.66.216) by m9.grp.scd.yahoo.com with QMQP;\n",
      "    22 Aug 2002 16:24:13 -0000\n",
      "Received: from unknown (HELO carbon) (194.73.73.92) by\n",
      "    mta1.grp.scd.yahoo.com with SMTP; 22 Aug 2002 16:24:15 -0000\n",
      "Received: from host217-36-23-185.in-addr.btopenworld.com ([217.36.23.185])\n",
      "    by carbon with esmtp (Exim 3.22 #8) id 17huko-0000JF-00 for\n",
      "    forteana@yahoogroups.com; Thu, 22 Aug 2002 17:24:14 +0100\n",
      "X-Mailer: Microsoft Outlook Express Macintosh Edition - 4.5 (0410)\n",
      "To: zzzzteana@yahoogroups.com\n",
      "X-Priority: 3\n",
      "Message-Id: <E17huko-0000JF-00@carbon>\n",
      "From: \"Tim Chapman\" <timc@2ubh.com>\n",
      "X-Yahoo-Profile: tim2ubh\n",
      "MIME-Version: 1.0\n",
      "Mailing-List: list zzzzteana@yahoogroups.com; contact\n",
      "    forteana-owner@yahoogroups.com\n",
      "Delivered-To: mailing list zzzzteana@yahoogroups.com\n",
      "Precedence: bulk\n",
      "List-Unsubscribe: <mailto:zzzzteana-unsubscribe@yahoogroups.com>\n",
      "Date: Thu, 22 Aug 2002 17:23:28 +0100\n",
      "Subject: Re: [zzzzteana] Which Muppet Are You?\n",
      "Reply-To: zzzzteana@yahoogroups.com\n",
      "Content-Type: text/plain; charset=US-ASCII\n",
      "Content-Transfer-Encoding: 7bit\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "> Apols if this has been posted before:\n",
      "> \n",
      "> http://www.pinkpaperclips.net/subs/quiz2.html\n",
      ">\n",
      "So, anyone who isn't Beaker?\n",
      "\n",
      "TimC\n",
      "Meep\n",
      "\n",
      "------------------------ Yahoo! Groups Sponsor ---------------------~-->\n",
      "4 DVDs Free +s&p Join Now\n",
      "http://us.click.yahoo.com/pt6YBB/NXiEAA/mG3HAA/7gSolB/TM\n",
      "---------------------------------------------------------------------~->\n",
      "\n",
      "To unsubscribe from this group, send an email to:\n",
      "forteana-unsubscribe@egroups.com\n",
      "\n",
      " \n",
      "\n",
      "Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ \n",
      "\n",
      "\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "header, subject, message = get_header_subject_message(X_train[20])\n",
    "print(subject)\n",
    "print('~-'*30)\n",
    "print(header)\n",
    "print('~-'*30)\n",
    "print(message)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_dom_cont_char(header):\n",
    "    pattern_d = {'dom_': 'From .+@(\\S+)', 'cont_': 'Content-Type: ([\\w/]+)\\;', 'char_': 'charset=\"?([\\w\\-\\_]+)\"?'}\n",
    "    get_results_d = {}\n",
    "    for key, value in pattern_d.items():\n",
    "        try_match = re.findall(value, header)\n",
    "        if len(try_match) != 0:\n",
    "            get_results_d[key] = try_match[0]\n",
    "        else:\n",
    "            get_results_d[key] = ' '\n",
    "    return get_results_d.values()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2ubh.com text/plain US-ASCII\n"
     ]
    }
   ],
   "source": [
    "dom, cont, char = get_dom_cont_char(header)\n",
    "print (dom, cont, char)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "masking_url = lambda x: re.sub('\\S*www\\S+|http://\\S+', 'urlmasked', x)\n",
    "masking_number = lambda x: re.sub('\\d+(?:[\\s\\,]?\\d+)*(?:\\.\\d+)?', 'digitnumber', x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "> Apols if this has been posted before:\n",
      "> \n",
      "> http://www.pinkpaperclips.net/subs/quiz2.html\n",
      ">\n",
      "So, anyone who isn't Beaker?\n",
      "\n",
      "TimC\n",
      "Meep\n",
      "\n",
      "------------------------ Yahoo! Groups Sponsor ---------------------~-->\n",
      "4 DVDs Free +s&p Join Now\n",
      "http://us.click.yahoo.com/pt6YBB/NXiEAA/mG3HAA/7gSolB/TM\n",
      "---------------------------------------------------------------------~->\n",
      "\n",
      "To unsubscribe from this group, send an email to:\n",
      "forteana-unsubscribe@egroups.com\n",
      "\n",
      " \n",
      "\n",
      "Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ \n",
      "\n",
      "\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(message)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "> Apols if this has been posted before:\n",
      "> \n",
      "> urlmasked\n",
      ">\n",
      "So, anyone who isn't Beaker?\n",
      "\n",
      "TimC\n",
      "Meep\n",
      "\n",
      "------------------------ Yahoo! Groups Sponsor ---------------------~-->\n",
      "digitnumber DVDs Free +s&p Join Now\n",
      "urlmasked\n",
      "---------------------------------------------------------------------~->\n",
      "\n",
      "To unsubscribe from this group, send an email to:\n",
      "forteana-unsubscribe@egroups.com\n",
      "\n",
      " \n",
      "\n",
      "Your use of Yahoo! Groups is subject to urlmasked \n",
      "\n",
      "\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(masking_number(masking_url(message)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Let's build the preprocessing class with these functions**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MessageSelector(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, keep_header=False, keep_subject=False, mask_number=True, mask_URL=True, add_main_par=True):\n",
    "        self.keep_header = keep_header\n",
    "        self.keep_subject = keep_subject\n",
    "        self.mask_number = mask_number\n",
    "        self.mask_URL = mask_URL\n",
    "        self.add_main_par = add_main_par\n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    def transform(self, X, y=None):\n",
    "        ind = X.index\n",
    "        X_mess = pd.Series(index=ind)\n",
    "        for ix in ind:\n",
    "            header, subject, content = get_header_subject_message(X[ix])\n",
    "            domain, cont_type, character = get_dom_cont_char(header)\n",
    "            main_sending_par = domain + ' ' + cont_type + ' ' + character\n",
    "            if self.mask_URL:\n",
    "                content, subject = masking_url(content), masking_url(subject)\n",
    "            if self.mask_number:\n",
    "                content, subject, header = masking_number(content), masking_number(subject), masking_number(header)\n",
    "            if self.add_main_par & (not self.keep_header): #want to keep usefull par. while removing header\n",
    "                content += '\\n' + main_sending_par\n",
    "            mail = self.keep_header * header + '\\n' + self.keep_subject * subject + '\\n' + content\n",
    "            X_mess[ix] = mail\n",
    "        return X_mess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Re: [zzzzteana] Which Muppet Are You?\n",
      "> Apols if this has been posted before:\n",
      "> \n",
      "> urlmasked\n",
      ">\n",
      "So, anyone who isn't Beaker?\n",
      "\n",
      "TimC\n",
      "Meep\n",
      "\n",
      "------------------------ Yahoo! Groups Sponsor ---------------------~-->\n",
      "digitnumber DVDs Free +s&p Join Now\n",
      "urlmasked\n",
      "---------------------------------------------------------------------~->\n",
      "\n",
      "To unsubscribe from this group, send an email to:\n",
      "forteana-unsubscribe@egroups.com\n",
      "\n",
      " \n",
      "\n",
      "Your use of Yahoo! Groups is subject to urlmasked \n",
      "\n",
      "\n",
      "\n",
      "\n",
      "2ubh.com text/plain US-ASCII\n"
     ]
    }
   ],
   "source": [
    "X_mess = MessageSelector(keep_subject=True).fit_transform(X_train)\n",
    "print(X_mess[20])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally we need to convert text into understandable values for a model. For that we will use the CountVectorizer class than will perform word tokenizing and occurence counting. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4836, 87075)\n",
      "77776\n",
      "17786\n"
     ]
    }
   ],
   "source": [
    "count_vect = CountVectorizer()\n",
    "X_train_count = count_vect.fit_transform(X_mess)\n",
    "print(X_train_count.shape)\n",
    "print(count_vect.vocabulary_.get('urlmasked'))\n",
    "print(count_vect.vocabulary_.get('digitnumber'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Going further we can also add a Tfidf transformer to compute an ajusted word occurence frequency upon counting."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Building the ML model**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Determining the short-list of most promising models**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
      "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
      "           weights='uniform')\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   24.5s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with knn : mean : 94.4%, Std : 0.76%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
      "          intercept_scaling=1, max_iter=100, multi_class='warn',\n",
      "          n_jobs=None, penalty='l2', random_state=None, solver='warn',\n",
      "          tol=0.0001, verbose=0, warm_start=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   20.5s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with logit : mean : 96.5%, Std : 0.54%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
      "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
      "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=None,\n",
      "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
      "       power_t=0.5, random_state=42, shuffle=True, tol=None,\n",
      "       validation_fraction=0.1, verbose=0, warm_start=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   20.4s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with sgd : mean : 98.4%, Std : 0.42%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=None,\n",
      "            oob_score=False, random_state=42, verbose=0, warm_start=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:  1.1min finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with rf : mean : 97.1%, Std : 0.57%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " LinearSVC(C=1.0, class_weight=None, dual=True, fit_intercept=True,\n",
      "     intercept_scaling=1, loss='hinge', max_iter=1000, multi_class='ovr',\n",
      "     penalty='l2', random_state=None, tol=0.0001, verbose=0)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   21.3s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with lin_svc : mean : 98.3%, Std : 0.37%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,\n",
      "  decision_function_shape='ovr', degree=3, gamma='auto', kernel='rbf',\n",
      "  max_iter=-1, probability=False, random_state=None, shrinking=True,\n",
      "  tol=0.001, verbose=False)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   59.1s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Accuracy - 5Fcv with rbf_svc : mean : 68.6%, Std : 0.03%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n",
      "\n",
      " MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)\n",
      "\n",
      " Accuracy - 5Fcv with nb_m : mean : 86.3%, Std : 0.94%\n",
      "~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-~-\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   19.4s finished\n"
     ]
    }
   ],
   "source": [
    "algo_d = {\n",
    "    'knn': KNeighborsClassifier(),\n",
    "    'logit': LogisticRegression(),\n",
    "    'sgd': SGDClassifier(random_state=42),\n",
    "    'rf': RandomForestClassifier(n_estimators=500, random_state=42),\n",
    "    'lin_svc': LinearSVC(loss='hinge'),\n",
    "    'rbf_svc': SVC(gamma='auto'),\n",
    "    'nb_m': MultinomialNB()\n",
    "}\n",
    "for key, clf in algo_d.items():\n",
    "    spam_clf = Pipeline([\n",
    "        ('mess_selector', MessageSelector()),\n",
    "        ('vectorizer', CountVectorizer()),\n",
    "        ('tfidf', TfidfTransformer()),\n",
    "        ('clf', clf)\n",
    "    ])\n",
    "    print(\"\\n\", clf)\n",
    "    cv = cross_val_score(spam_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, scoring='accuracy')\n",
    "    print(\"\\n Accuracy - 5Fcv with {} : mean : {:.1f}%, Std : {:.2f}%\".format(\n",
    "        key, 100 * cv.mean(), 100 * np.std(cv)))\n",
    "    print('~-'*30)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Linear SVC sounds very promising ! Its stochastic \"variant\" here gives even slightly better results (randomness might help getting out of local optima ?) ! On the opposite, rbf svc acts like a dumb classifier, assuming everything is ham e-mail (most frequent target class).<br> <br>\n",
    "Multinomial Naive Bayes classifier is a bit disappointing. It's known to work very well for text analysis. Multinomial NB classifier is suitable with discret features (like word counts) and multinomial distribution requires normal integer features counts but, in practice, fractional counts (like the ones provided by tf-idf) also work. Let's give a try with count vectors instead of fractionnal ones. I will also give a chance to bernoulli NB classifier with 0-1 counts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   19.1s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy - 5Fcv with Mnb : mean : 94.4%, Std : 1.35%\n",
      "Accuracy - 5Fcv with Bnb : mean : 86.8%, Std : 1.09%\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   18.9s finished\n"
     ]
    }
   ],
   "source": [
    "Mnb_clf = Pipeline([\n",
    "    ('mess_selector', MessageSelector()),\n",
    "    ('vectorizer', CountVectorizer()),\n",
    "    ('Mnb', MultinomialNB())\n",
    "])\n",
    "\n",
    "cv = cross_val_score(Mnb_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, scoring='accuracy')\n",
    "print(\"Accuracy - 5Fcv with Mnb : mean : {:.1f}%, Std : {:.2f}%\".format(100 * cv.mean(), 100 * np.std(cv)))\n",
    "\n",
    "Bnb_clf = Pipeline([\n",
    "    ('mess_selector', MessageSelector()),\n",
    "    ('vectorizer', CountVectorizer(binary=True)),\n",
    "    ('Bnb', BernoulliNB())\n",
    "])\n",
    "cv = cross_val_score(Bnb_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, scoring='accuracy')\n",
    "print(\"Accuracy - 5Fcv with Bnb : mean : {:.1f}%, Std : {:.2f}%\".format(100 * cv.mean(), 100 * np.std(cv)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Much better for the Multinomial NB clf ! <br> <br>\n",
    "Although accuracy is not so terrible as a metric here (considering the rather balanced distribution of target classes, roughly 65 VS 35) but we should rely on a more robust metric including recall and precision. Let's have a look at precision-recall curves for our best classifiers at this stage of the study. I will keep SGD with hinge loss for linear SVC algo (no predict proba method for Linear SVC implying to use SVC class or probability calibration)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   19.3s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   18.9s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   20.2s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:   19.4s finished\n",
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:  1.0min finished\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_score_d = {}\n",
    "\n",
    "sgd_clf = Pipeline([\n",
    "    ('mess_selector', MessageSelector()),\n",
    "    ('vectorizer', CountVectorizer()),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('sgd', SGDClassifier(random_state=42))\n",
    "])\n",
    "y_score_sgd = cross_val_predict(sgd_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, method='decision_function')\n",
    "y_score_d['sgd'] = y_score_sgd\n",
    "\n",
    "y_proba_Mnb = cross_val_predict(Mnb_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, method='predict_proba')\n",
    "y_score_Mnb = y_proba_Mnb[:, 1]\n",
    "y_score_d['Mnb'] = y_score_Mnb\n",
    "\n",
    "\n",
    "algo_short_d = {\n",
    "    'knn': KNeighborsClassifier(),\n",
    "    'logit': LogisticRegression(),\n",
    "    'rf': RandomForestClassifier(n_estimators=500, random_state=42)\n",
    "}\n",
    "for key, clf in algo_short_d.items():\n",
    "    spam_clf = Pipeline([\n",
    "        ('mess_selector', MessageSelector()),\n",
    "        ('vectorizer', CountVectorizer()),\n",
    "        ('tfidf', TfidfTransformer()),\n",
    "        ('clf', clf)\n",
    "    ])\n",
    "    y_proba = cross_val_predict(spam_clf, X_train, y_train, cv=5, verbose=2, n_jobs=-1, method='predict_proba')\n",
    "    y_score = y_proba[:, 1]\n",
    "    y_score_d[key] = y_score\n",
    "    \n",
    "plt.figure(figsize=(10, 8))\n",
    "for key, score in y_score_d.items():\n",
    "    precisions, recalls, thresholds = precision_recall_curve(y_train, score)\n",
    "    plt.plot(recalls, precisions, label=key)\n",
    "plt.ylim([0, 1.01])\n",
    "plt.xlabel('Recall')\n",
    "plt.ylabel('Precision')\n",
    "plt.legend(loc='center')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SGD for linear SVC, Random Forests and Logit give the most promising results. Logit and SGD for linear SVC might be a bit a redundant here (linear svc should out-perform the logit anyway ...). I will only carry on with SGD and RF for this exercise. Extra spare time for computation could then be allocated to search the preprocessing parameter space. In a real-life study, LinearSVC classifier should also be investigated and one might even want to try mini-batch gradient descent !"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Fine-tuning the models**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "*First, let's add a lemmatizer or stemmer that might slightly improve the models* \n",
    "<br>\n",
    "As mentionned in Sklearn documentation : \"Fancy token-level analysis such as stemming, lemmatizing, compound splitting, filtering based on part-of-speech, etc. are not included in the scikit-learn codebase, but can be added by customizing either the tokenizer or the analyzer\".\n",
    "Let's build custom classes for a token-lemmatizer/stemmer using NLTK."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "class StemTokenizer(object):\n",
    "    def __init__(self, stm=EnglishStemmer()):\n",
    "        self.stm = stm\n",
    "    def __call__(self, doc): \n",
    "        return [self.stm.stem(t) for t in word_tokenize(doc)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "class LemmaTokenizer(object):\n",
    "    def __init__(self, wnl=WordNetLemmatizer()):\n",
    "        self.wnl = wnl\n",
    "    def __call__(self, doc):\n",
    "        return [self.wnl.lemmatize(t) for t in word_tokenize(doc)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These classes can be modified by replacing word_tokenize(doc) to CountVectorizer().build_analyzer()(doc)\n",
    "<br>\n",
    "In this case the custom class should be pass as an analyzer within CountVectorizer(). Here I prefered to keep hands on punctuation for further investigations and therefore choose to pass the custom class to tokenizer."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'where': 21, 'were': 20, 'your': 22, 'cat': 3, 'go': 7, '?': 1, 'they': 18, 'head': 8, 'to': 19, 'the': 16, 'northern': 11, 'part': 14, 'of': 12, 'citi': 4, '.': 0, 'and': 2, 'my': 10, 'did': 5, 'it': 9, 'onc': 13, 'feel': 6, 'stronger': 15, 'then': 17}\n",
      "{'where': 21, 'were': 20, 'your': 22, 'cat': 3, 'going': 7, '?': 1, 'they': 18, 'headed': 8, 'to': 19, 'the': 16, 'northern': 11, 'part': 14, 'of': 12, 'city': 4, '.': 0, 'and': 2, 'yours': 23, 'my': 10, 'did': 5, 'it': 9, 'once': 13, 'feel': 6, 'stronger': 15, 'then': 17}\n"
     ]
    }
   ],
   "source": [
    "corpus = [\"Where were your cats going ?\",\n",
    "          \"They headed to the northern part of the city. And yours ?\",\n",
    "          \"My cat did it once. It feels stronger then\"]\n",
    "\n",
    "stem_vectorizer = CountVectorizer(tokenizer=StemTokenizer())\n",
    "stem_vectorizer.fit_transform(corpus)\n",
    "print(stem_vectorizer.vocabulary_)\n",
    "lem_vectorizer = CountVectorizer(tokenizer=LemmaTokenizer())\n",
    "lem_vectorizer.fit_transform(corpus)\n",
    "print(lem_vectorizer.vocabulary_)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's add a final option to our previous MessageSelector class to handle custom punctuation selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MessagePunctSelector(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, keep_header=False, keep_subject=False, add_main_par=True, keep_punct=False,\n",
    "                 mask_number=True, mask_URL=True):\n",
    "        self.keep_header = keep_header\n",
    "        self.keep_subject = keep_subject\n",
    "        self.mask_number = mask_number\n",
    "        self.mask_URL = mask_URL\n",
    "        self.add_main_par = add_main_par\n",
    "        self.keep_punct = keep_punct\n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    def transform(self, X, y=None):\n",
    "        ind = X.index\n",
    "        X_mess = pd.Series(index=ind)\n",
    "        for ix in ind:\n",
    "            header, subject, content = get_header_subject_message(X[ix])\n",
    "            domain, cont_type, character = get_dom_cont_char(header)\n",
    "            main_sending_par = domain + ' ' + cont_type + ' ' + character\n",
    "            if self.mask_URL:\n",
    "                content, subject = masking_url(content), masking_url(subject)\n",
    "            if self.mask_number:\n",
    "                content, subject, header = masking_number(content), masking_number(subject), masking_number(header)\n",
    "            if not self.keep_punct:\n",
    "                text_punct = '''!\"#%&'()*+,-./:;<=>?@[\\]^_`{|}~''' #want to keep $ for 'money' occurence\n",
    "                header_punct = '''!\"#$%&'()*+,:;<=>?@[\\]^`{|}~''' #specific punctuation for header\n",
    "                translator_text = str.maketrans(text_punct, ' '*len(text_punct)) #prefer empty space to None\n",
    "                translator_header = str.maketrans(header_punct, ' '*len(header_punct)) #prefer empty space to None\n",
    "                content, subject = content.translate(translator_text), subject.translate(translator_text)\n",
    "                header = header.translate(translator_header)\n",
    "            if self.add_main_par & (not self.keep_header):\n",
    "                content += '\\n' + main_sending_par\n",
    "            mail = self.keep_header * header + '\\n' + self.keep_subject * subject + '\\n' + content\n",
    "            X_mess[ix] = mail\n",
    "        return X_mess"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### **Hyperparameter tuning**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Considering the large number of hyperparameters that could be modified throughout the preprocessing steps and the computation time required, fine tuning will be sliced in three consecutive GridsearchCV works. It's not the best approach but it should work just fine for this exercise. If extra time, another strategies could be to use a big parameter distribution and perform a large randomize search or to conduct the search in two steps : the first roughly determining best approximate buckets for parameters and the second to refine the first search. <br> <br>\n",
    "\n",
    "Here I decided to firstly tune all the preprocessing parameters and then to look for the best hyperparameters of the clf.<br>\n",
    "For metric I will go with the 'average_precision' score, corresponding to area under the precision-recall curve (equivalent to auc score for ROC curves, up to 1 and the higher the better !). Depending on final purpose, precision or recall would also have been relevant."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *Linear SVC with stochastic gradient descent*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 96 candidates, totalling 288 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed: 40.3min\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed: 143.4min\n",
      "[Parallel(n_jobs=-1)]: Done 288 out of 288 | elapsed: 173.2min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('mess', MessagePunctSelector(add_main_par=True, keep_header=False, keep_punct=False,\n",
       "           keep_subject=False, mask_URL=True, mask_number=True)), ('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='...dom_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], 'mess__mask_number': [True, False], 'mess__keep_punct': [True, False], 'vect__lowercase': [True], 'vect__tokenizer': [<__main__.StemTokenizer object at 0x00000258E366FA20>, <__main_...enizer': [None], 'vect__token_pattern': ['\\\\w\\\\w+', '\\\\w\\\\w+|[!\"#%&\\'()*+,-./:;<=>?@[\\\\]^_`{|}~]']}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf = Pipeline([\n",
    "    ('mess', MessagePunctSelector()),\n",
    "    ('vect', CountVectorizer()),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('sgd', SGDClassifier(random_state=42))\n",
    "])\n",
    "param_grid = [\n",
    "    {'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], \n",
    "     'mess__mask_number': [True, False], 'mess__keep_punct': [True, False], 'vect__lowercase': [True],\n",
    "     'vect__tokenizer': [StemTokenizer(), LemmaTokenizer()]\n",
    "     },\n",
    "    {'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], \n",
    "     'mess__mask_number': [True, False], 'mess__keep_punct': [True], 'vect__lowercase': [True], \n",
    "     'vect__tokenizer': [None], 'vect__token_pattern': ['\\w\\w+', '''\\w\\w+|[!\"#%&'()*+,-./:;<=>?@[\\]^_`{|}~]''']\n",
    "     } #specific custom pattern for tokenization with/wo punctuation\n",
    "]\n",
    "cv = GridSearchCV(sgd_clf, param_grid=param_grid, cv=3, verbose=2, n_jobs=-1, scoring='average_precision')\n",
    "cv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.82%, best parameters : {'mess__keep_header': True, 'mess__keep_punct': False, 'mess__keep_subject': False, 'mess__mask_URL': False, 'mess__mask_number': True, 'vect__lowercase': True, 'vect__tokenizer': <__main__.StemTokenizer object at 0x00000258E366FA20>}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 27 candidates, totalling 81 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed: 47.0min\n",
      "[Parallel(n_jobs=-1)]: Done  81 out of  81 | elapsed: 107.2min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('mess', MessagePunctSelector(add_main_par=True, keep_header=True, keep_punct=False,\n",
       "           keep_subject=False, mask_URL=False, mask_number=True)), ('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='...dom_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'vect__ngram_range': [(1, 1), (1, 2), (2, 2)], 'vect__max_df': [1.0, 0.8, 0.6], 'vect__min_df': [1, 0.001, 0.005]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=False, keep_punct=False, \n",
    "                                  mask_URL=False, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=StemTokenizer())),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('sgd', SGDClassifier(random_state=42))\n",
    "])\n",
    "param_grid = [\n",
    "    {'vect__ngram_range': [(1, 1), (1, 2), (2, 2)], 'vect__max_df': [1.0, 0.8, 0.6], \n",
    "     'vect__min_df': [1, 0.001, 0.005]\n",
    "     }\n",
    "]\n",
    "cv = GridSearchCV(sgd_clf, param_grid=param_grid, cv=3, verbose=2, n_jobs=-1, scoring='average_precision')\n",
    "cv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.86%, best parameters : {'vect__max_df': 0.6, 'vect__min_df': 0.001, 'vect__ngram_range': (1, 2)}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If available sparte time, you might want to investigate further max_df parameter for CountVectorizer..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "prep_pip = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=False, keep_punct=False, \n",
    "                                  mask_URL=False, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=StemTokenizer(), \n",
    "                             max_df=0.6, min_df=0.001, ngram_range=(1, 2))),\n",
    "    ('tfidf', TfidfTransformer())\n",
    "])\n",
    "\n",
    "X_prep = prep_pip.fit_transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 288 candidates, totalling 864 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:  1.7min\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed:  5.0min\n",
      "[Parallel(n_jobs=-1)]: Done 357 tasks      | elapsed: 18.8min\n",
      "[Parallel(n_jobs=-1)]: Done 640 tasks      | elapsed: 40.4min\n",
      "[Parallel(n_jobs=-1)]: Done 864 out of 864 | elapsed: 58.1min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=None,\n",
       "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "       power_t=0.5, random_state=42, shuffle=True, tol=None,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'alpha': array([1.e-03, 1.e-04, 1.e-05]), 'penalty': ['l2'], 'loss': ['hinge', 'squared_hinge'], 'early_stopping': [True, False], 'n_iter_no_change': [5], 'learning_rate': ['optimal', 'adaptive'], 'eta0': [0.05], 'max_iter': [5, 206, 1000]}, {'alpha': array([1.e-03, 1.e-04, 1.e-05]), 'p...change': [5], 'learning_rate': ['optimal', 'adaptive'], 'eta0': [0.05], 'max_iter': [5, 206, 1000]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf = SGDClassifier(random_state=42)\n",
    "\n",
    "param_grid = [\n",
    "    {'alpha': 10.0**-np.arange(3, 6), 'penalty': [\"l2\"], 'loss': [\"hinge\", \"squared_hinge\"], \n",
    "     'early_stopping': [True, False], 'n_iter_no_change':[5], 'learning_rate': ['optimal', 'adaptive'], \n",
    "     'eta0': [0.05], 'max_iter':[5, 10**6//len(X_train), 1000]},\n",
    "    {'alpha': 10.0**-np.arange(3, 6), 'penalty': [\"elasticnet\"], 'loss': [\"hinge\", \"squared_hinge\"], \n",
    "     'l1_ratio': [0.15, 0.25, 0.4], 'early_stopping': [True, False], 'n_iter_no_change':[5],\n",
    "     'learning_rate': ['optimal', 'adaptive'],  'eta0': [0.05], 'max_iter':[5, 10**6//len(X_train), 1000]}\n",
    "]\n",
    "cv = GridSearchCV(sgd_clf, param_grid=param_grid, scoring='average_precision', cv=3, verbose=2, n_jobs=-1)\n",
    "cv.fit(X_prep, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.88%, best parameters : {'alpha': 1e-05, 'early_stopping': False, 'eta0': 0.05, 'learning_rate': 'adaptive', 'loss': 'hinge', 'max_iter': 206, 'n_iter_no_change': 5, 'penalty': 'l2'}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:  4.8min finished\n"
     ]
    }
   ],
   "source": [
    "sgd_clf_final = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=False, keep_punct=False, \n",
    "                                  mask_URL=False, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=StemTokenizer(), \n",
    "                             max_df=0.6, min_df=0.001, ngram_range=(1, 2))),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('sgd', SGDClassifier(random_state=42, alpha=1e-05, loss='hinge', max_iter=10**6//len(X_train),\n",
    "                          penalty='l2', early_stopping=False, learning_rate='adaptive', n_iter_no_change=5,\n",
    "                          eta0=0.05))\n",
    "])\n",
    "y_pred = cross_val_predict(sgd_clf_final, X_train, y_train, cv=5, verbose=2, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy score on validation set: 98.92%\n",
      "Final precision score on validation set: 98.10%\n",
      "Final recall score on validation set: 98.48%\n"
     ]
    }
   ],
   "source": [
    "print(\"Final accuracy score on validation set: {:.2f}%\".format(accuracy_score(y_train, y_pred)*100))\n",
    "print(\"Final precision score on validation set: {:.2f}%\".format(precision_score(y_train, y_pred)*100))\n",
    "print(\"Final recall score on validation set: {:.2f}%\".format(recall_score(y_train, y_pred)*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   3 out of   3 | elapsed:  2.4min finished\n"
     ]
    }
   ],
   "source": [
    "y_score = cross_val_predict(sgd_clf_final, X_train, y_train, cv=3, verbose=2, n_jobs=-1, \n",
    "                            method='decision_function')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x432 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax_arr = plt.subplots(1, 2, figsize=(16, 6))\n",
    "precisions, recalls, thresholds = precision_recall_curve(y_train, y_score)\n",
    "ax_arr[0].plot(recalls, precisions, label='Final model with SGD classifier')\n",
    "ax_arr[0].set(xlabel='Recall', ylabel='Precision', ylim=[0.8, 1.01], xlim=[0, 1.01])\n",
    "ax_arr[1].plot(thresholds, precisions[:-1], \"b--\", label='Precision')\n",
    "ax_arr[1].plot(thresholds, recalls[:-1], \"g-\", label='Recall')\n",
    "ax_arr[1].set(xlabel='Thresholds', ylim=[0.8, 1.01])\n",
    "ax_arr[1].legend(loc='best')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Threshold value can still be changed to aim at a specific precision/recall tradeoff. For example, let's try to reach a precision of 99%. Looking at the precision-recall/tresholds curves, the corresponding thresholds should be arount 0.3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "New precision score on validation set: 99.06%\n",
      "New recall score on validation set: 96.84%\n"
     ]
    }
   ],
   "source": [
    "y_pred_99 = (y_score > 0.3)\n",
    "print(\"New precision score on validation set: {:.2f}%\".format(precision_score(y_train, y_pred_99)*100))\n",
    "print(\"New recall score on validation set: {:.2f}%\".format(recall_score(y_train, y_pred_99)*100))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "29 23\n"
     ]
    }
   ],
   "source": [
    "FP_train = X_train[(y_train == 0) & (y_pred == 1)].reset_index(drop=True)\n",
    "FN_train = X_train[(y_train == 1) & (y_pred == 0)].reset_index(drop=True)\n",
    "print(len(FP_train), len(FN_train))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Return-Path: <mailer@desktopengineer.com>\n",
      "Delivered-To: Jost.Krieger+freetechmail@ruhr-uni-bochum.de\n",
      "Received: (qmail 8473 invoked by alias); 24 Jul 2002 00:13:25 -0000\n",
      "Received: (qmail 8465 invoked by uid 82); 24 Jul 2002 00:13:25 -0000\n",
      "Received: from mailer@desktopengineer.com by mailhost with qmail-scanner-1.00 (uvscan: v4.1.40/v4213. . Clean. Processed in 0.632524 secs); 24 Jul 2002 00:13:25 -0000\n",
      "Received: from enterprisedesktop.com (HELO server1.enterprisedesktop.com) (209.61.182.209)\n",
      "  by mi-1.rz.ruhr-uni-bochum.de with SMTP; 24 Jul 2002 00:13:24 -0000\n",
      "Received: from localhost.localdomain (localhost [127.0.0.1])\n",
      "\tby server1.enterprisedesktop.com (8.11.3/8.11.3) with SMTP id g6O0D2k32482\n",
      "\tfor <Jost.Krieger+freetechmail@ruhr-uni-bochum.de>; Tue, 23 Jul 2002 20:13:02 -0400\n",
      "Message-Id: <200207240013.g6O0D2k32482@server1.enterprisedesktop.com>\n",
      "Received: from Octeth Email Manager Pro ([207.34.72.13]) by www.desktopengineer.com with HTTP (Apache/1.3.22 (Unix)  (Red-Hat/Linux) mod_ssl/2.8.5 OpenSSL/0.9.6 PHP/4.0.4pl1 mod_perl/1.21 FrontPage/4.0.4.3);\n",
      "\t Tue, 23 Jul 2002 20:13:02 -0400\n",
      "Date: Tue, 23 Jul 2002 20:13:02 -0400\n",
      "To: \"Jost.Krieger+freetechmail@ruhr-uni-bochum.de \" <Jost.Krieger+freetechmail@ruhr-uni-bochum.de>\n",
      "From: \"DesktopEngineer.com\" <mailer@desktopengineer.com>\n",
      "Reply-to: \"Jost.Krieger+freetechmail@ruhr-uni-bochum.de \" <Jost.Krieger+freetechmail@ruhr-uni-bochum.de>\n",
      "Subject: [DEJD] DesktopEngineer.com Headlines - $200 For The Best Submission\n",
      "X-Priority: 3\n",
      "X-Sender: DesktopEngineer.com <mailer@desktopengineer.com>\n",
      "X-Mailer: Octeth Email Manager Pro\n",
      "X-MSMail-Priority: Medium\n",
      "Importance: Medium\n",
      "MIME-Version: 1.0\n",
      "Content-Type: Multipart/Mixed;\n",
      "\tboundary=\"Boundary-=_bcea19c9d6337c63081baf1d5ee3466aa\"\n",
      "\n",
      "This is a MIME message. If you are reading this text, you\n",
      "might want to consider changing to a mail reader that\n",
      "understands how to properly display MIME multipart messages.\n",
      "\n",
      "--Boundary-=_bcea19c9d6337c63081baf1d5ee3466aa\n",
      "Content-Type: multipart/alternative; charset = \"iso-8859-1\";\n",
      "\tboundary=\"Boundary-=_sb73baa252922041f1e5a038a67575b2a2\";\n",
      "\n",
      "--Boundary-=_sb73baa252922041f1e5a038a67575b2a2\n",
      "Content-Type: text/plain; charset = \"iso-8859-1\";\n",
      "Content-Transfer-Encoding: 8bit\n",
      "\n",
      "If you are seeing this message, please visit http://desktopengineer.com to see the latest stories.\n",
      "\n",
      "--Boundary-=_sb73baa252922041f1e5a038a67575b2a2\n",
      "Content-Type: text/html; charset = \"iso-8859-1\";\n",
      "Content-Transfer-Encoding: 8bit\n",
      "\n",
      "<!DOCTYPE HTML PUBLIC \"-//SoftQuad Software//DTD HoTMetaL PRO 6.0::19990601::extensions to HTML 4.0//EN\" \"hmpro6.dtd\">\n",
      "<HTML> \n",
      "  <HEAD> \n",
      "\t <TITLE>Desktop Engineers Junk Drawer - Handy Odds N Ends for Windows 98, NT\n",
      "\t\tand 2000</TITLE> <META NAME=\"GeekLog\" CONTENT=\"Created with GeekLog 1.5rc1\">\n",
      "\t <META HTTP-EQUIV=\"expires\" CONTENT=\"60\"> <STYLE TYPE=\"text/css\">\n",
      "<!--\t\n",
      "BODY, TD\t{ color: black; font-size: x-small; font-family: verdana, arial, sans-serif }\n",
      "H1 \t\t{ color: black; font-size: medium; font-weight: bold; font-family: verdana, arial, sans-serif }\n",
      "H2 \t\t{ color: black; font-size: small; font-weight: bold; font-family: verdana, arial, sans-serif }\n",
      ".blocktitle \t{ color: white; font-size: x-small; font-weight: bold; font-family: verdana, arial, sans-serif }\n",
      ".footer \t{ font-size: xx-small; font-family: verdana, arial, sans-serif }\n",
      "-->\n",
      "</STYLE> \n",
      "  </HEAD> \n",
      "  <BODY TEXT=\"#000000\" BGCOLOR=\"#FFFFFF\" LINK=\"#333333\" VLINK=\"#333333\"\n",
      "\tALINK=\"#333333\" LEFTMARGIN=\"0\" TOPMARGIN=\"0\">\n",
      "<!-- begin header.tmpl -->\n",
      "\t <TABLE BORDER=\"0\" WIDTH=\"630\" ALIGN=\"LEFT\" CELLSPACING=\"0\" CELLPADDING=\"0\">\n",
      "\t\t\n",
      "\t\t<TR ALIGN=\"right\" VALIGN=\"middle\"> \n",
      "\t\t  <TD COLSPAN=\"2\" HEIGHT=\"18\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR VALIGN=\"bottom\"> \n",
      "\t\t  <TD COLSPAN=\"2\" ALIGN=\"LEFT\" WIDTH=\"630\"><A\n",
      "\t\t\t HREF=\"http://desktopengineer.com\" TARGET=\"promo\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/deblue5.gif\" HEIGHT=\"84\" BORDER=\"0\"\n",
      "\t\t\t ALT=\"http://desktopengineer.com\" WIDTH=\"630\"></A></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\" BGCOLOR=\"#000000\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"1\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\" BGCOLOR=\"#AAAAAA\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"6\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\" BGCOLOR=\"#FFFFFF\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"1\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR BGCOLOR=\"#648DAD\"> \n",
      "\t\t  <TD COLSPAN=\"2\"><B><FONT COLOR=\"#333333\">Desktop Engineer's Junk Drawer\n",
      "\t\t\t Headline Mailer!</FONT></B></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\" BGCOLOR=\"#AAAAAA\" HEIGHT=\"6\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\" BGCOLOR=\"#EEEEEE\"><SMALL>To subscribe, send a blank\n",
      "\t\t\t email to: <A\n",
      "\t\t\t HREF=\"mailto:dejdweekly-subscribe@topica.com\">dejdweekly-subscribe@topica.com</A></SMALL></TD>\n",
      "\t\t  \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\"> <BR> \n",
      "\t\t\t <TABLE CELLSPACING=\"0\" CELLPADDING=\"1\" WIDTH=\"500\" BORDER=\"0\"\n",
      "\t\t\t ALIGN=\"CENTER\"> \n",
      "\t\t\t\t<TBODY> \n",
      "\t\t\t\t  <TR BGCOLOR=\"#304a63\"> \n",
      "\t\t\t\t\t <TD> \n",
      "\t\t\t\t\t\t<TABLE CELLSPACING=\"0\" CELLPADDING=\"2\" WIDTH=\"100%\"\n",
      "\t\t\t\t\t\tBORDER=\"0\"> \n",
      "\t\t\t\t\t\t  <TBODY> \n",
      "\t\t\t\t\t\t\t <TR BGCOLOR=\"#304a63\"> \n",
      "\t\t\t\t\t\t\t\t<TD CLASS=\"blocktitle\" COLSPAN=\"2\"\n",
      "\t\t\t\t\t\t\t\tALIGN=\"CENTER\">ADVERTISEMENT<BR><FONT COLOR=\"#FF0000\"> Discover the Powerful\n",
      "\t\t\t\t\t\t\t\t  Repackaging Capabilities of AdminStudio</FONT></TD> \n",
      "\t\t\t\t\t\t\t </TR> \n",
      "\t\t\t\t\t\t\t <TR> \n",
      "\t\t\t\t\t\t\t\t<TD BGCOLOR=\"#e9edf2\" COLSPAN=\"2\">Let InstallShield\n",
      "\t\t\t\t\t\t\t\t  show you the easy process for converting packages to MSI by demonstrating an\n",
      "\t\t\t\t\t\t\t\t  application conversion using AdminStudio 3. Get detailed explanations of many\n",
      "\t\t\t\t\t\t\t\t  best practices, and don't miss the Q&amp;A session afterwards. Register Now!\n",
      "\t\t\t\t\t\t\t\t  <A\n",
      "\t\t\t\t\t\t\t\t\tHREF=\"http://www.installshield.com/0207/DesktopEngineer0723\">http://www.installshield.com/0207/DesktopEngineer0723</A></TD>\n",
      "\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t\t </TR> \n",
      "\t\t\t\t\t\t  </TBODY> \n",
      "\t\t\t\t\t\t</TABLE></TD> \n",
      "\t\t\t\t  </TR> \n",
      "\t\t\t\t</TBODY> \n",
      "\t\t\t </TABLE><BR></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD><IMG SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"10\"\n",
      "\t\t\t HEIGHT=\"1\" ALT=\"\"></TD> \n",
      "\t\t  <TD> <HR><B>DesktopEngineer.com Giving Away $200 Submission Of The\n",
      "\t\t\t Month Award</B><BR> Friday, July 12 2002 @ 07:41 AM EDT by Darwin_Sanoy<BR> \n",
      "\t\t\t <P>How about a little cash to buy something really cool? Send us\n",
      "\t\t\t\tthose tips! </P>\n",
      "\t\t\t <A\n",
      "\t\t\t  HREF=\"http://desktopengineer.com/article.php?story=20020718005048357\">http://desktopengineer.com/article.php?story=20020718005048357</A><BR><BR>\n",
      "\t\t\t <HR><B>Microsoft Webcast: WMIC: Managing Windows from a Command Line (07/24/02\n",
      "\t\t\t @ 10am PST)</B><BR> Thursday, July 11 2002 @ 11:22 AM EDT by Darwin_Sanoy<BR> \n",
      "\t\t\t <P>WMI is the powerful management API Microsoft built into Windows\n",
      "\t\t\t\t2000. Getting scripted access to this management functionality can not be done\n",
      "\t\t\t\tthrough shell scripts - until WMIC was released with XP. </P>\n",
      "\t\t\t <A\n",
      "\t\t\t  HREF=\"http://desktopengineer.com/article.php?story=20020718002155930\">http://desktopengineer.com/article.php?story=20020718002155930</A><BR><BR>\n",
      "\t\t\t <HR> \n",
      "\t\t\t <P ALIGN=\"CENTER\">\n",
      "\t\t\t\t<A\n",
      "\t\t\t\t HREF=\"http://desktopengineer.com/cgi-bin/ads/adclick.cgi?manager=adcycle.com&gid=1&cid=26&mid=84&id=472\"\n",
      "\t\t\t\t TARGET=\"_blank\"><IMG\n",
      "\t\t\t\tSRC=\"http://desktopengineer.com/cgi-bin/ads/adcycle.cgi?gid=1&cid=26&mid=84&id=472\"\n",
      "\t\t\t\tWIDTH=\"468\" HEIGHT=\"60\" BORDER=\"0\" ALT=\"Click to Visit InstallShield\"></A></P>\n",
      "\t\t\t <HR><B>Wise Black Belt Series: Application Isolation Webinar</B><BR> Wednesday,\n",
      "\t\t\t July 10 2002 @ 06:12 AM EDT by Darwin_Sanoy<BR> \n",
      "\t\t\t <P><B>Session: Application Isolation and Windows Installer</B> This\n",
      "\t\t\t\twebinar covered how .LOCAL and .NET (Manifests) isolation work. Webinar\n",
      "\t\t\t\tincluded demonstrations of both isolation technologies for traditional\n",
      "\t\t\t\tapplications. </P>\n",
      "\t\t\t <A\n",
      "\t\t\t  HREF=\"http://desktopengineer.com/article.php?story=2002071415404212\">http://desktopengineer.com/article.php?story=2002071415404212</A><BR><BR>\n",
      "\t\t\t <HR><B>Nice Linksys 4 Port USB Mini-Hub for $11.65 at Buy.com</B><BR> Tuesday,\n",
      "\t\t\t July 09 2002 @ 07:55 AM EDT by Darwin_Sanoy<BR> \n",
      "\t\t\t <P>This is a great prices for four ports! </P>\n",
      "\t\t\t <A\n",
      "\t\t\t  HREF=\"http://desktopengineer.com/article.php?story=20020718005709995\">http://desktopengineer.com/article.php?story=20020718005709995</A><BR><BR><HR>\n",
      "\t\t\t </TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD COLSPAN=\"2\"> <BR> \n",
      "\t\t\t <TABLE CELLSPACING=\"0\" CELLPADDING=\"1\" WIDTH=\"500\" BORDER=\"0\"\n",
      "\t\t\t  ALIGN=\"CENTER\"> \n",
      "\t\t\t\t<TBODY> \n",
      "\t\t\t\t  <TR BGCOLOR=\"#304a63\"> \n",
      "\t\t\t\t\t <TD> \n",
      "\t\t\t\t\t\t<TABLE CELLSPACING=\"0\" CELLPADDING=\"2\" WIDTH=\"100%\"\n",
      "\t\t\t\t\t\t BORDER=\"0\"> \n",
      "\t\t\t\t\t\t  <TBODY> \n",
      "\t\t\t\t\t\t\t <TR BGCOLOR=\"#304a63\"> \n",
      "\t\t\t\t\t\t\t\t<TD CLASS=\"blocktitle\" COLSPAN=\"2\"\n",
      "\t\t\t\t\t\t\t\t ALIGN=\"CENTER\">ADVERTISEMENT<BR><FONT COLOR=\"#FF0000\"> Discover the Powerful\n",
      "\t\t\t\t\t\t\t\t  Repackaging Capabilities of AdminStudio</FONT></TD> \n",
      "\t\t\t\t\t\t\t </TR> \n",
      "\t\t\t\t\t\t\t <TR> \n",
      "\t\t\t\t\t\t\t\t<TD BGCOLOR=\"#e9edf2\" COLSPAN=\"2\">Let InstallShield\n",
      "\t\t\t\t\t\t\t\t  show you the easy process for converting packages to MSI by demonstrating an\n",
      "\t\t\t\t\t\t\t\t  application conversion using AdminStudio 3. Get detailed explanations of many\n",
      "\t\t\t\t\t\t\t\t  best practices, and don't miss the Q&amp;A session afterwards. Register Now!\n",
      "\t\t\t\t\t\t\t\t  <A\n",
      "\t\t\t\t\t\t\t\t  HREF=\"http://www.installshield.com/0207/DesktopEngineer0723\">http://www.installshield.com/0207/DesktopEngineer0723</A></TD>\n",
      "\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t\t </TR> \n",
      "\t\t\t\t\t\t  </TBODY> \n",
      "\t\t\t\t\t\t</TABLE></TD> \n",
      "\t\t\t\t  </TR> \n",
      "\t\t\t\t</TBODY> \n",
      "\t\t\t </TABLE><BR> </TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD BGCOLOR=\"#000000\" COLSPAN=\"2\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"1\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD BGCOLOR=\"#AAAAAA\" COLSPAN=\"2\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"6\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD BGCOLOR=\"#EEEEEE\" COLSPAN=\"2\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"1\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR BGCOLOR=\"#648DAD\" VALIGN=\"top\"> \n",
      "\t\t  <TD HEIGHT=\"20\" CLASS=\"footer\" COLSPAN=\"2\">&nbsp;Copyright &copy; 2002\n",
      "\t\t\t Desktop Engineers Junk Drawer, ISSN: 1522-7693<BR> &nbsp;All trademarks and\n",
      "\t\t\t copyrights on this page are owned by their respective owners.</TD> \n",
      "\t\t</TR> \n",
      "\t\t<TR> \n",
      "\t\t  <TD BGCOLOR=\"#AAAAAA\" COLSPAN=\"2\"><IMG\n",
      "\t\t\t SRC=\"http://desktopengineer.com/images/speck.gif\" WIDTH=\"1\" HEIGHT=\"1\"\n",
      "\t\t\t ALT=\"\"></TD> \n",
      "\t\t</TR> \n",
      "\t </TABLE> </BODY>\n",
      "</HTML>\n",
      "\n",
      "\n",
      "--Boundary-=_sb73baa252922041f1e5a038a67575b2a2--\n",
      "\n",
      "\n",
      "--Boundary-=_bcea19c9d6337c63081baf1d5ee3466aa--\n",
      "\n",
      "\n",
      " Return-Path: nobody@example.com\n",
      "Delivery-Date: Wed Sep 18 00:00:00 1972\n",
      "Received: from timhunt.net (ns1.timhunt.net [216.27.147.130])\n",
      "\tby dogma.slashnull.org (8.11.6/8.11.6) with ESMTP id g6Q4AC423036\n",
      "\tfor <tim_timhunt_net@spamtraps.taint.org>; Fri, 26 Jul 2002 05:10:13 +0100\n",
      "Received: from applenews.lists.apple.com (applenews.lists.apple.com [17.254.0.152])\n",
      "\tby timhunt.net (8.11.3/8.11.1) with ESMTP id g6Q49SE45947\n",
      "\tfor <sagfsg@fsg.com>; Fri, 26 Jul 2002 00:09:28 -0400 (EDT)\n",
      "\t(envelope-from owner-quicktime_news@applenews.lists.apple.com)\n",
      "Received: from localhost (daemon@localhost)\n",
      "        by applenews.lists.apple.com (8.11.6/8.11.6) with SMTP id g6Q3QAX10809\n",
      "        ; Thu, 25 Jul 2002 20:26:10 -0700 (PDT)\n",
      "Date: Thu, 25 Jul 2002 20:26:10 -0700 (PDT)\n",
      "Errors-to: owner-quicktime_news@applenews.lists.apple.com\n",
      "Received: by applenews.lists.apple.com (ulkbay_ailermay v1.5); Thu, 25 Jul 2002 20:13:25 -0700\n",
      "To: QuickTime News <quicktime_news@applenews.lists.apple.com>\n",
      "From: QuickTime News <quicktime_news@applenews.lists.apple.com>\n",
      "Subject: QuickTime News: July 26, 2002\n",
      "MIME-Version: 1.0\n",
      "Content-Type: multipart/alternative; charset=us-ascii;\n",
      "\tboundary=\"This_is_the_boundary_hip_ho\"\n",
      "Content-Transfer-Encoding: 7bit\n",
      "Precedence: bulk\n",
      "Message-ID: <bulk.1889.20020725201325@applenews.lists.apple.com>\n",
      "\n",
      "\n",
      "--This_is_the_boundary_hip_ho\n",
      "Content-Type: text/plain; charset=us-ascii\n",
      "Content-Transfer-Encoding: 7bit\n",
      "\n",
      "You appear to be using an email application that won't properly\n",
      "display the graphical (or HTML) version of our newsletter. Because\n",
      "we want you to enjoy reading QuickTime News, we recommend that you\n",
      "visit the following site and switch to the plain text version of\n",
      "this newsletter:\n",
      "\n",
      "    http://www.apple.com/enews/change/\n",
      "\n",
      "If you have trouble reading today's issue of QuickTime News, please\n",
      "consider visiting the following web page where it has been posted\n",
      "for your enjoyment:\n",
      "\n",
      "    http://www.apple.com/enews/currentqissue/\n",
      "    \n",
      "If you receive a blank message with an enclosure, drag the enclosure\n",
      "into the window of your web browser and today's issue of QuickTime News\n",
      "will display in your web browser.\n",
      "\n",
      "Thank you.\n",
      "\n",
      "\n",
      "--This_is_the_boundary_hip_ho\n",
      "Content-Type: text/html; charset=us-ascii\n",
      "Content-Transfer-Encoding: 7bit\n",
      "\n",
      "\n",
      "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.01 Transitional//EN\">\n",
      "<html>\n",
      "<head>\n",
      "\t<title>QuickTime News: July 26, 2002</title>\n",
      "\t<meta name=\"generator\" content=\"BBEdit 6.0.1\">\n",
      "</head>\n",
      "<body background=\"http://a772.g.akamai.net/7/772/51/c241f0d493a981/www.apple.com/enews/eimages/stripes.gif\" bgcolor=\"#CCCCCC\">\n",
      "\n",
      "<!-- START MAIN TABLE -->\n",
      "<a name=\"top\"></a>\n",
      "\n",
      "<table border=\"0\" width=\"600\" cellspacing=\"0\" cellpadding=\"0\" align=\"center\">\n",
      "\t<tr>\n",
      "\t\t<td><img src=\"http://a772.g.akamai.net/7/772/51/9fa804f4688ffc/www.apple.com/enews/2002/05/images/box-topleft.gif\" alt=\"\" width=\"1\" height=\"14\" border=\"0\"></td>\n",
      "\t\t<td><img src=\"http://a772.g.akamai.net/7/772/51/c22ed219eeb953/www.apple.com/enews/2002/05/images/box-topmiddle.gif\" alt=\"\" width=\"598\" height=\"14\" border=\"0\"></td>\n",
      "\t\t<td><img src=\"http://a772.g.akamai.net/7/772/51/97518749dde4f5/www.apple.com/enews/2002/05/images/box-topright.gif\" alt=\"\" width=\"1\" height=\"14\" border=\"0\"></td>\n",
      "\t</tr>\n",
      "\t<tr>\n",
      "\t\t<td bgcolor=\"#7e7e7e\"><img src=\"http://a772.g.akamai.net/7/772/51/84485f130fd0fe/www.apple.com/enews/eimages/spacer.gif\" alt=\"\" width=\"1\" height=\"1\" border=\"0\"></td>\n",
      "\t\t<td bgcolor=\"#ffffff\">\n",
      "\t\t<a href=\"http://www.apple.com/quicktime/\"><img src=\"http://a772.g.akamai.net/7/772/51/8c1280c5572c24/www.apple.com/enews/2002/05/images/logo-quicktime.gif\" alt=\"\" width=\"366\" height=\"52\" border=\"0\"></a><a href=\"http://www.apple.com/enews/schedule/\"><img src=\"http://a772.g.akamai.net/7/772/51/405c7d71bf1a3b/www.apple.com/enews/2002/07/images/26vol.gif\" alt=\"\" width=\"232\" height=\"52\" border=\"0\"></a><br>\n",
      "\t\t<a href=\"http://www.apple.com/hotnews/\"><img src=\"http://a1552.g.akamai.net/7/1552/51/1e05f706dc3a33/www.apple.com/enews/2002/07/images/25ticker.gif\" alt=\"\" width=\"598\" height=\"28\" border=\"0\"></a><br><br>\n",
      "\t\t\n",
      "\t\t<table border=\"0\" cellpadding=\"0\" cellspacing=\"0\" width=\"598\">\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td><img src=\"http://a772.g.akamai.net/7/772/51/84485f130fd0fe/www.apple.com/enews/eimages/spacer.gif\" alt=\"\" width=\"8\" height=\"1\" border=\"0\"></td>\n",
      "\t\t\t\t<td valign=\"top\">\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/93e5c294690bb5/www.apple.com/enews/eimages/thisissue.gif\" alt=\"In This Issue:\" width=\"150\" height=\"40\" border=\"0\"><br>\n",
      "\t\t\t\t<a href=\"#article1\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/bb9fb46de41f5f/www.apple.com/enews/2002/07/images/26qtc1.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article2\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/f8e16e24427062/www.apple.com/enews/2002/07/images/26qtc2.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article3\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/95c481448e9005/www.apple.com/enews/2002/07/images/26qtc3.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article4\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/67f3124e242d18/www.apple.com/enews/2002/07/images/26qtc4.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article5\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/9562dd6341274e/www.apple.com/enews/eimages/toc_qtpark.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article1\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/7cee5fe5dea417/www.apple.com/enews/eimages/toc_qtmusic.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t<a href=\"#article7\"><img src=\"http://a772.g.akamai.net/7/772/51/3efcf8160adc84/www.apple.com/enews/eimages/toc_arrow.gif\" alt=\"\" width=\"11\" height=\"20\" border=\"0\"><img src=\"http://a772.g.akamai.net/7/772/51/5de80697882f02/www.apple.com/enews/eimages/toc_quicktakes.gif\" alt=\"\" width=\"340\" height=\"20\" border=\"0\"></a><br>\n",
      "\t\t\t\t</td>\n",
      "\t\t\t\t<td>\n",
      "\t\t\t\t<a href=\"http://www.apple.com/trailers/wb/matrix_reloaded.html\"><img src=\"http://a772.g.akamai.net/7/772/51/f12e1402a6f4b5/www.apple.com/enews/2002/07/images/matrix2.jpg\" alt=\"\" width=\"160\" height=\"132\" border=\"0\"></a>\n",
      "\t\t\t\t</td>\n",
      "\t\t\t</tr>\n",
      "\t\t</table>\n",
      "\t\t<br>\n",
      "\t\t\n",
      "\t\t<table border=\"0\" cellpadding=\"0\" cellspacing=\"0\" width=\"598\">\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#e5e5e5\" colspan=\"3\"><img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"598\" height=\"1\" border=\"0\"></td>\n",
      "\t\t\t</tr>\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#ffffff\" valign=\"top\" width=\"377\">\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/84485f130fd0fe/www.apple.com/enews/eimages/spacer.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 1 --><a name=\"article1\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/05d9778050e4e9/www.apple.com/enews/2002/07/images/26qh1.gif\" alt=\"\" width=\"344\" height=\"27\" border=\"0\"></a><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/hotnews/articles/2002/07/mwny/\"><img src=\"http://a772.g.akamai.net/7/772/51/78c70455db9bfe/www.apple.com/enews/2002/07/images/nycab.jpg\" alt=\"\" width=\"155\" height=\"121\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tOn July 17, tens of thousands packed New York City&#146;s Jacob K. Javitz\n",
      "\t\t\t\t\t\tCenter&#151;and more than 50,000 tuned in online &#151; to watch Apple CEO\n",
      "\t\t\t\t\t\tSteve Jobs unveil exciting new Apple products and technologies at\n",
      "\t\t\t\t\t\tMacworld.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tMiss the keynote? Not to worry, we already have an <a href=\"http://www.apple.com/quicktime/qtv/mwny02/\">MPEG-4 stream</a>\n",
      "\t\t\t\t\t\tavailable on our website.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tWhat can you expect to see?\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tWatch Jobs highlight the amazing new features&#151;there are over\n",
      "\t\t\t\t\t\t150 of them&#151;in Jaguar, Mac OS X v10.2, due to ship in late August.\n",
      "\t\t\t\t\t\tLearn about the new benefits of Apple&#146;s online .Mac service. And\n",
      "\t\t\t\t\t\tcheck out the new 17-inch widescreen iMac, as well as the newest\n",
      "\t\t\t\t\t\tiPod models&#151;including iPod for Windows.\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 2 --><a name=\"article2\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/de19f93c7c8fd5/www.apple.com/enews/2002/07/images/26qh2.gif\" alt=\"\" width=\"344\" height=\"27\" border=\"0\"></a><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/quicktime/products/qt/\"><img src=\"http://a772.g.akamai.net/7/772/51/b34cb1cb78c08a/www.apple.com/enews/2002/07/images/QTlogo.gif\" alt=\"\" width=\"100\" height=\"82\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tQuickTime 6 is now available for <a href=\"http://www.apple.com/quicktime/download/\">download</a>.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tIncorporating MPEG-4 and its next-generation digital media\n",
      "\t\t\t\t\t\ttechnology, QuickTime 6 delivers an unparalleled multimedia\n",
      "\t\t\t\t\t\texperience. It features:\n",
      "\t\t\t\t\t\t<br clear=\"all\"><br>\n",
      "\t\t\t\t\t\t<table border=\"0\" cellpadding=\"0\" cellspacing=\"0\">\n",
      "\t\t\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">&#149;&nbsp;&nbsp; </font></td>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">Instant-On streaming. No more waiting through buffer delays.\n",
      "\t\t\t\t\t\tInstant-On gives you instant access to whatever section of a movie\n",
      "\t\t\t\t\t\tyou want to jump to.</font></td>\n",
      "\t\t\t\t\t\t\t</tr>\n",
      "\t\t\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">&#149;&nbsp;&nbsp; </font></td>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">Advanced Audio Coding (AAC). The new standard in professional audio,\n",
      "\t\t\t\t\t\tAAC provides more efficient compression than ever yet delivers\n",
      "\t\t\t\t\t\tsound quality rivaling that of uncompressed CD audio.</font></td>\n",
      "\t\t\t\t\t\t\t</tr>\n",
      "\t\t\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">&#149;&nbsp;&nbsp; </font></td>\n",
      "\t\t\t\t\t\t\t\t<td valign=\"top\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">An updated user interface. Accessing QuickTime content is now easier\n",
      "\t\t\t\t\t\tthan ever. We&#146;ve also added a new Favorites menu, so you can bookmark\n",
      "\t\t\t\t\t\tyou favorite QuickTime movies.</font></td>\n",
      "\t\t\t\t\t\t\t</tr>\n",
      "\t\t\t\t\t\t</table>\n",
      "\t\t\t\t\t\t<br>\n",
      "\t\t\t\t\t\tAnd that&#146;s just the <a href=\"http://www.apple.com/quicktime/products/qt/\">beginning</a> of the great QuickTime 6 story.\n",
      "\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 3 --><a name=\"article3\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/fa075b8742884b/www.apple.com/enews/2002/07/images/26qh3.gif\" alt=\"\" width=\"344\" height=\"27\" border=\"0\"></a><br>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/ipod/\"><img src=\"http://a772.g.akamai.net/7/772/51/c69266217d00c8/www.apple.com/enews/2002/07/images/ipodwin.jpg\" alt=\"\" width=\"155\" height=\"165\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tWe know that QuickTime News readers love music. And soon, more of\n",
      "\t\t\t\t\t\tyou will be able to take your favorite tunes with you wherever you\n",
      "\t\t\t\t\t\tgo. That&#146;s because iPod, Apple&#146;s incredibly popular MP3 player, is\n",
      "\t\t\t\t\t\tnow available for both Mac and Windows users.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/ipod\">iPod for Windows</a> comes with the leading jukebox software for\n",
      "\t\t\t\t\t\tPCs&#151;MUSICMATCH Jukebox Plus. So you can easily manage your digital\n",
      "\t\t\t\t\t\tmusic library. Quickly transfer your music collection&#151;and\n",
      "\t\t\t\t\t\tplaylists&#151;to iPod via FireWire. Automatically sync with iPod when\n",
      "\t\t\t\t\t\tyour music library changes. And use it to convert your music CDs,\n",
      "\t\t\t\t\t\ttapes and vinyl to MP3s.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tiPod for Windows ships in August, but you can order yours today from\n",
      "\t\t\t\t\t\tthe Apple Store. \n",
      "\n",
      "\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 4 --><a name=\"article4\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/eb4a44cf4859cc/www.apple.com/enews/2002/07/images/26qh4.gif\" alt=\"\" width=\"344\" height=\"28\" border=\"0\"></a><br><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/quicktime/whatson/\"><img src=\"http://a772.g.akamai.net/7/772/51/ded24b6aeecb8f/www.apple.com/enews/2002/07/images/ozzy.jpg\" alt=\"\" width=\"130\" height=\"113\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tLike to find great QuickTime content all in one place? Visit our new\n",
      "\t\t\t\t\t\tstreamlined website. You&#146;ll now find links to a vast assortment of\n",
      "\t\t\t\t\t\tQuickTime video and audio consolidated onto our new <a href=\"http://www.apple.com/quicktime/whatson/\">What&#146;s On</a> page.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tTune in, for example, to Ozzy Osbourne. Enjoying new-found celebrity\n",
      "\t\t\t\t\t\tthanks to his hit cable series, Ozzy performs &#147;Junkie&#148; live at Budokan on Epic \n",
      "\t\t\t\t\t\tRecords&#146; New Music Mondays.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tOr explore the surreal world of realMyst. MacPlay&#146;s 3D version of\n",
      "\t\t\t\t\t\tthe classic game Myst takes you on a dreamlike journey of mystical\n",
      "\t\t\t\t\t\tadventure. Is it really just a game?\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tJump back into reality by visiting VRway. Over 3000 QuickTime VRs\n",
      "\t\t\t\t\t\ttake you to some of the most beautiful sites of Europe.\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\n",
      "\t\t\t\t<!-- Article 4b --><a name=\"article4\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/74d5945dcc6372/www.apple.com/enews/2002/05/images/03qh4.gif\" alt=\"\" width=\"344\" height=\"27\" border=\"0\"></a><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/quicktime/upgrade/\"><img src=\"http://a772.g.akamai.net/7/772/51/5d9f53de4124f7/www.apple.com/enews/2002/07/images/qtpro.jpg\" alt=\"\" width=\"140\" height=\"125\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tEager to author content in the new international standard format for\n",
      "\t\t\t\t\t\tthe Internet&#151;MPEG-4? Now you can do so easily by upgrading to\n",
      "\t\t\t\t\t\tQuickTime 6 Pro.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tWith QuickTime 6 Pro, not only can you harness the power of MPEG-4,\n",
      "\t\t\t\t\t\tbut you can play movies in full screen, edit movies by simply\n",
      "\t\t\t\t\t\tcutting and pasting, prepare movies for streaming.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tLearn more about <a href=\"http://www.apple.com/quicktime/upgrade/\">QuickTime 6 Pro</a> on our website.\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tFor those of you who want the ability to preview MPEG-2 content, we\n",
      "\t\t\t\t\t\tnow offer the <a href=\"http://www.apple.com/quicktime/products/mpeg2playback/\">QuickTime 6 MPEG-2 Playback Component</a>, available for\n",
      "\t\t\t\t\t\tpurchase at the online Apple Store.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tBy adding this component to QuickTime 6, you have the perfect tool\n",
      "\t\t\t\t\t\tfor previewing and sharing your work with your clients throughout\n",
      "\t\t\t\t\t\tthe production process.\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 5 --><a name=\"article5\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/b1ef1322d15e1c/www.apple.com/enews/2002/05/images/31qh6.gif\" alt=\"\" width=\"204\" height=\"27\" border=\"0\"></a><br>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/trailers/touchstone/signs/\"><img src=\"http://a772.g.akamai.net/7/772/51/fe5a07ea09f1a8/www.apple.com/enews/2002/07/images/signs.jpg\" alt=\"\" width=\"105\" height=\"150\" border=\"0\" align=\"right\"></a>\n",
      "\t\t\t\t\t\tIn rural Pennsylvania, pastor Graham Hess and his family are\n",
      "\t\t\t\t\t\tterrified when mysterious giant crop circles appear in their\n",
      "\t\t\t\t\t\tcornfield&#151;as well as around the world. Is it a sign from alien\n",
      "\t\t\t\t\t\tbeings? Or something else?\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tStarring Mel Gibson and directed by M. Night Shyamalan (&#147;The Sixth\n",
      "\t\t\t\t\t\tSense&#148;), &#147;<a href=\"http://www.apple.com/trailers/touchstone/signs/\">Signs</a>&#148; opens in U.S. theaters on August 2.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tSome believe that supernatural forces create crop circles, while\n",
      "\t\t\t\t\t\tothers remain skeptical. To help you decide, Touchstone Pictures\n",
      "\t\t\t\t\t\tpresents images of actual circles, as well as interviews with\n",
      "\t\t\t\t\t\twitnesses and experts. Click on &#147;<a href=\"http://bventertainment.go.com/movies/signs/\">Is it a Sign?</a>&#148; to find out more.\n",
      "\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"377\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\" width=\"100%\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td align=\"center\"><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tWe hope you enjoyed reading this issue of QuickTime News.<br>\n",
      "\t\t\t\t\t\tWe&#146;ll send your next issue on August 9.<br>\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\n",
      "\t\t\t\t</td>\n",
      "\t\t\t\t<td bgcolor=\"#e5e5e5\" width=\"1\"><img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"1\" height=\"1\" border=\"0\"></td>\n",
      "\t\t\t\t<td bgcolor=\"#ffffff\" valign=\"top\" width=\"220\">\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/84485f130fd0fe/www.apple.com/enews/eimages/spacer.gif\" alt=\"\" width=\"220\" height=\"1\" border=\"0\"><br>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 5 --><a name=\"article1\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/6f9427aa8f03c0/www.apple.com/enews/2002/05/images/03qh5.gif\" alt=\"\" width=\"204\" height=\"27\" border=\"0\"></a><br>\n",
      "\n",
      "\t\t\t\t\t\t<a href=\"http://getmusic.com/microsites/michaelbolton/\"><img src=\"http://a772.g.akamai.net/7/772/51/3d1a049d1ddfe2/www.apple.com/enews/2002/07/images/MBoltonCD.jpg\" alt=\"\" width=\"100\" height=\"100\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tAfter four years out of the spotlight, singer/songwriter Michael\n",
      "\t\t\t\t\t\tBolton returns with a new album, &#147;<a href=\"http://www.michaelboltonmusic.com\">Only a Woman Like You</a>.&#148; Watch the\n",
      "\t\t\t\t\t\tvideo for the first single (which bears the same name as the album),\n",
      "\t\t\t\t\t\tco-written by Bolton, country sensation Shania Twain, and her\n",
      "\t\t\t\t\t\thusband, record producer Mutt Lang.\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<a href=\"http://www.bethorton.mu\"><img src=\"http://a772.g.akamai.net/7/772/51/6e4cdf6a36f414/www.apple.com/enews/2002/07/images/bethorton.jpg\" alt=\"\" border=\"0\" align=\"right\"></a>\n",
      "\t\t\t\t\t\tBeth Orton&#151;the British singer/songwriter whose eagerly awaited third album arrives \n",
      "\t\t\t\t\t\ton US store shelves next week&#151;invited a few vocal friends to join her on &#147;<a href=\"http://www.bethorton.mu\">Daybreaker</a>.&#148;\n",
      "\t\t\t\t\t\tLike Emmylou Harris (who chimes in on &#147;God Song&#148;), the Chemical Brothers, Ben Watt, Johnny Marr and William Orbit.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tVisit Orton&#146;s site and watch the music <i>video</i> for &#147;Concrete Sky&#148; and hear \n",
      "\t\t\t\t\t\tthree <i>audio</i> tracks from the album: &#147;Carmella,&#148; &#147;Daybreaker,&#148; and &#147;Thinking \n",
      "\t\t\t\t\t\tAbout Tomorrow.&#148;\n",
      "\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\n",
      "\t\t\t\t\t\tWhat&#146;s new on Columbia&#146;s Midweek Music?\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\t<a href=\"http://columbiarecords.com/midweekmusic/\"><img src=\"http://a772.g.akamai.net/7/772/51/d27baaa466600d/www.apple.com/enews/2002/07/images/aerials.jpg\" alt=\"\" width=\"100\" height=\"86\" border=\"0\" align=\"right\"></a>\n",
      "\t\t\t\t\t\tOne of the headlining bands at this year&#146;s Ozzfest, L.A.&#146;s\n",
      "\t\t\t\t\t\tmulticultural metal men System of a Down perform &#147;<a href=\"http://columbiarecords.com/midweekmusic/\">Aerials</a>,&#148; the\n",
      "\t\t\t\t\t\tlatest from their 2001 album &#147;Toxicity.&#148; Watch the video through\n",
      "\t\t\t\t\t\tJuly 31.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tAnd from <i>July 31 to August 7</i>, check out &#147;<a href=\"http://columbiarecords.com/midweekmusic/\">Girls of Summer</a>,&#148; the\n",
      "\t\t\t\t\t\tvideo for one of two new tracks included on the classic compilation,\n",
      "\t\t\t\t\t\t&#147;O Yeah! Ultimate Aerosmith Hits.&#148;\n",
      "\t\t\t\t\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t&#147;They&#146;re personal/political odes with the bite of twenty\n",
      "\t\t\t\t\t\trottweilers. Passionate songs full of fear and loathing, but with a\n",
      "\t\t\t\t\t\tpop rush.&#148; So says the REM-inspired, London band Wilt about their\n",
      "\t\t\t\t\t\tnew album, &#147;My Medicine.&#148;\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tReady for a dose? Then tune in to hear <a href=\"http://www.w-mm.co.uk\">eleven complete tracks</a>.\n",
      "\t\t\t\t\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<a href=\"http://www.rhino.com/fun/listeningparties/78184_PartyPlayer.mgi2?P=RHIQT\"><img src=\"http://a772.g.akamai.net/7/772/51/59514d2e33f301/www.apple.com/enews/2002/07/images/RandyTravis.jpg\" alt=\"\" width=\"100\" height=\"90\" border=\"0\" align=\"left\"></a>\n",
      "\t\t\t\t\t\tIn 1986, with his debut album &#147;Storms of Life,&#148; Randy Travis became\n",
      "\t\t\t\t\t\tthe first country music artist to go multi-platinum. &#147;Trail of\n",
      "\t\t\t\t\t\tMemories: The Randy Travis Anthology&#148; brings together over 40 of the\n",
      "\t\t\t\t\t\tlegendary star&#146;s hits from 1986 to 1999, including duets with George\n",
      "\t\t\t\t\t\tJones, B.B. King, Tammy Wynette, and Merle Haggard. Join the <a href=\"http://www.rhino.com/fun/listeningparties/78184_PartyPlayer.mgi2?P=RHIQT\">Randy\n",
      "\t\t\t\t\t\tTravis Listening Party</a> and enjoy Disc One of this 2-disc collection.\n",
      "\t\t\t\t\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tThe music of alternative rock band <a href=\"http://www.swither.com/\">Swither</a> evokes both the opacity\n",
      "\t\t\t\t\t\tand richness of human emotions. Watch the MPEG-4 videos of &#147;Here and\n",
      "\t\t\t\t\t\tNow&#148; and &#147;Pri-Mer.&#148;\n",
      "\t\t\t\t\t\t\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"220\" height=\"1\" border=\"0\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<!-- Article 7 --><a name=\"article7\"></a>\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font size=2 face=\"Lucida Grande,Geneva,Arial,Helvetica,sans-serif\" class=\"L12\">\n",
      "\t\t\t\t\t\t<a href=\"#top\"><img src=\"http://a772.g.akamai.net/7/772/51/39bcd94d38b883/www.apple.com/enews/2002/05/images/02ih8.gif\" alt=\"\" width=\"204\" height=\"27\" border=\"0\"></a><br>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t&#147;So much music, it makes me feel powerful.&#148; That&#146;s <a href=\"http://www.apple.com/hotnews/articles/2002/07/mwny/hamiltonmorris.html\">Hamilton Morris</a>\n",
      "\t\t\t\t\t\ttalking about his iPod in one of the new ads featured in last\n",
      "\t\t\t\t\t\tweek&#146;s Macworld New York expo. Catch the QuickTime movie.\n",
      "\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/3dfa419404e68b/www.apple.com/enews/2002/07/images/tiefly.jpg\" alt=\"\" width=\"100\" height=\"76\" border=\"0\" align=\"right\">\n",
      "\t\t\t\t\t\tHow good are your fly-tying techniques? Watch these tutorials from\n",
      "\t\t\t\t\t\tabout.com and learn how fly fishermen tie one on the right way.\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tFirst up, watch <a href=\"http://flyfishing.about.com/library/ffm/ftb/aajbwhipfinish.htm\">QuickTime video seminars</a> illustrating how to tie the\n",
      "\t\t\t\t\t\tDouble Hitch and Hand Whip-Finish and how to use a Matarelli\n",
      "\t\t\t\t\t\tWhip-finish Tool.\n",
      "\t\t\t\t\t\t<br><br>\n",
      "\t\t\t\t\t\tThen, in <a href=\"http://flyfishing.miningco.com/library/ffm/skills/aajbknotvideos.htm\">another video seminar</a>, find out how to tie the Improved\n",
      "\t\t\t\t\t\tClinch Knot, Non-Slip Loop, George Harvey Dry-Fly Knot, and\n",
      "\t\t\t\t\t\tNail-less Nail Knot.\n",
      "\n",
      "\t\t\t\t\t\t<p align=\"center\">\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/82815b4b56f654/www.apple.com/enews/eimages/separator.jpg\" alt=\"\" width=\"33\" height=\"12\" border=\"0\" align=\"middle\">\t\t\t\t\n",
      "\t\t\t\t\t\t</p>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tIf you&#146;re dying to know what the dead relatives of famous people are\n",
      "\t\t\t\t\t\tthinking, investigate the celebrity readings on psychic John\n",
      "\t\t\t\t\t\tEdward&#146;s <a href=\"http://www.scifi.com/johnedward/celebrity/\">Crossing Over</a> website.\n",
      "\n",
      "\t\t\t\t\t\t</font>\n",
      "\t\t\t\t\t\t</td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\n",
      "\t\t\t\t</td>\n",
      "\t\t\t</tr>\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#e5e5e5\" colspan=\"3\"><img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"598\" height=\"1\" border=\"0\"></td>\n",
      "\t\t\t</tr>\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#ffffff\" colspan=\"3\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td><font face=\"Geneva, Verdana, Arial, Helvetica, sans-serif\" size=\"1\" class=\"G10\">\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tQuickTime News is a free, bi-weekly email publication.<br><br>\n",
      "\t\t\n",
      "\t\t\t\t\t\tEvent dates are subject to change. Some products, programs, or promotions are not available outside the U.S. Visit your local \n",
      "\t\t\t\t\t\tApple site or call your local authorized Apple reseller for more information. Prices are estimated retail prices and are listed \n",
      "\t\t\t\t\t\tin U.S. dollars. Product specifications are subject to change.\t\t\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font></td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t</td>\n",
      "\t\t\t</tr>\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#e5e5e5\" colspan=\"3\"><img src=\"http://a772.g.akamai.net/7/772/51/358f94896524e7/www.apple.com/enews/2002/05/images/greypixel.gif\" alt=\"\" width=\"598\" height=\"1\" border=\"0\"></td>\n",
      "\t\t\t</tr>\n",
      "\t\t\t<tr>\n",
      "\t\t\t\t<td bgcolor=\"#ffffff\" colspan=\"3\" align=\"center\">\n",
      "\t\t\t\t\n",
      "\t\t\t\t<table border=\"0\" cellpadding=\"8\" cellspacing=\"0\">\n",
      "\t\t\t\t\t<tr>\n",
      "\t\t\t\t\t\t<td align=\"center\"><font face=\"Geneva, Verdana, Arial, Helvetica, sans-serif\" size=\"1\" class=\"G10\">\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t<img src=\"http://a772.g.akamai.net/7/772/51/c219cc85bc9fbf/www.apple.com/enews/eimages/red_apple_logo.gif\" width=\"25\" height=\"30\" border=\"0\" alt=\"\"><br>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\tCopyright&#169;2002 Apple Computer, Inc.&nbsp;&nbsp;<a href=\"http://www.apple.com/legal/default.html\">All Rights Reserved</a><br>\n",
      "\t\t\t\t\t\t<a href=\"http://www.apple.com/enews/subscribe/\">Subscribe</a> | <a href=\"http://www.apple.com/enews/unsubscribe/\">Unsubscribe</a> | <a href=\"http://www.apple.com/enews/change/\">Change Settings</a> | <a href=\"http://www.apple.com/legal/privacy/\">Privacy Policy</a></font>\n",
      "\t\t\t\t\t\t\n",
      "\t\t\t\t\t\t</font></td>\n",
      "\t\t\t\t\t</tr>\n",
      "\t\t\t\t</table>\n",
      "\t\t\t\t\n",
      "\t\t\t\t</td>\n",
      "\t\t\t</tr>\n",
      "\t\t</table>\n",
      "\t\t\n",
      "\t\t</td>\t\t\n",
      "\t\t<td bgcolor=\"#7e7e7e\"><img src=\"http://a772.g.akamai.net/7/772/51/84485f130fd0fe/www.apple.com/enews/eimages/spacer.gif\" alt=\"\" width=\"1\" height=\"1\" border=\"0\"></td>\n",
      "\t</tr>\t\t\n",
      "\t<tr>\n",
      "\t\t<td colspan=\"3\"><img src=\"http://a772.g.akamai.net/7/772/51/aa230f243aeaa0/www.apple.com/enews/2002/05/images/box-bottom.gif\" alt=\"\" width=\"600\" height=\"12\" border=\"0\"></td>\n",
      "\t</tr>\n",
      "</table>\n",
      "\n",
      "\n",
      "</body>\n",
      "\n",
      "\n",
      "\n",
      "\n",
      "--This_is_the_boundary_hip_ho--\n",
      "\n",
      "\n",
      " Return-Path: ler@lerami.lerctr.org\n",
      "Delivery-Date: Mon Sep  9 08:44:17 2002\n",
      "Received: via dmail-2002(12) for +lists/pgsql/sql; Mon, 9 Sep 2002 08:44:17 -0500 (CDT)\n",
      "Return-Path: <pgsql-sql-owner+M9320@postgresql.org>\n",
      "Received: from relay1.pgsql.com (relay1.pgsql.com [64.49.215.129])\n",
      "\tby lerami.lerctr.org (8.12.2/8.12.2/20020902/$Revision: 1.30 $) with ESMTP id g89DiBE9024795\n",
      "\tfor <ler@lerctr.org>; Mon, 9 Sep 2002 08:44:12 -0500 (CDT)\n",
      "Received: from postgresql.org (postgresql.org [64.49.215.8])\n",
      "\tby relay1.pgsql.com (Postfix) with ESMTP\n",
      "\tid 98ADA743D68; Mon,  9 Sep 2002 09:43:52 -0400 (EDT)\n",
      "Received: from localhost (postgresql.org [64.49.215.8])\n",
      "\tby postgresql.org (Postfix) with ESMTP id 413114763CA\n",
      "\tfor <pgsql-sql@postgresql.org>; Mon,  9 Sep 2002 04:57:48 -0400 (EDT)\n",
      "Received: from addr-mx02.addr.com (addr-mx02.addr.com [209.249.147.146])\n",
      "\tby postgresql.org (Postfix) with ESMTP id EE82C4763BE\n",
      "\tfor <pgsql-sql@postgresql.org>; Mon,  9 Sep 2002 04:57:46 -0400 (EDT)\n",
      "Received: from proxy1.addr.com (proxy1.addr.com [209.249.147.28])\n",
      "\tby addr-mx02.addr.com (8.12.5/8.12.2) with ESMTP id g898vlSV060755\n",
      "\tfor <pgsql-sql@postgresql.org>; Mon, 9 Sep 2002 01:57:47 -0700 (PDT)\n",
      "Received: from JD1 ([202.71.153.170])\n",
      "\tby proxy1.addr.com (8.11.6/8.9.1) with SMTP id g898vks91718\n",
      "\tfor <pgsql-sql@postgresql.org>; Mon, 9 Sep 2002 01:57:46 -0700 (PDT)\n",
      "\t(envelope-from jayps@addr.com)(envelope-to <pgsql-sql@postgresql.org>)\n",
      "Message-ID: <005101c257df$0343be50$a800a8c0@blraddrcom>\n",
      "From: \"Jay\" <jayps@addr.com>\n",
      "To: <pgsql-sql@postgresql.org>\n",
      "Subject: [SQL] Database joins\n",
      "Date: Mon, 9 Sep 2002 14:28:04 +0530\n",
      "MIME-Version: 1.0\n",
      "X-Priority: 3\n",
      "X-MSMail-Priority: Normal\n",
      "X-Mailer: Microsoft Outlook Express 6.00.2600.0000\n",
      "X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000\n",
      "X-Scanned-By: MIMEDefang 2.15 (www dot roaringpenguin dot com slash mimedefang)\n",
      "X-Virus-Scanned: by amavisd-milter (http://amavis.org/)\n",
      "X-Virus-Scanned: by AMaViS new-20020517\n",
      "Precedence: bulk\n",
      "Sender: pgsql-sql-owner@postgresql.org\n",
      "Content-Type: multipart/alternative; boundary=\"----=_NextPart_000_004E_01C2580D.1BC61260\"\n",
      "X-Status: \n",
      "X-Keywords: \n",
      "\n",
      "This is a multi-part message in MIME format.\n",
      "\n",
      "------=_NextPart_000_004E_01C2580D.1BC61260\n",
      "Content-Type: text/plain;\n",
      "\tcharset=\"iso-8859-1\"\n",
      "Content-Transfer-Encoding: quoted-printable\n",
      "\n",
      "Hi,\n",
      "I have created two databases, db1 and db2  in my POSTGRESQL database system.\n",
      "And both the databases contains few tables also.\n",
      "Could you please help me to write a query which should retrieve data from b=\n",
      "oth the databases\n",
      "(Database joins and table joins)\n",
      "Expecting and earliest reply\n",
      "Regards\n",
      "Jay\n",
      "\n",
      "\n",
      "------=_NextPart_000_004E_01C2580D.1BC61260\n",
      "Content-Type: text/html;\n",
      "\tcharset=\"iso-8859-1\"\n",
      "Content-Transfer-Encoding: quoted-printable\n",
      "\n",
      "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.0 Transitional//EN\">\n",
      "<HTML><HEAD>\n",
      "<META http-equiv=3DContent-Type content=3D\"text/html; charset=3Diso-8859-1\">\n",
      "<META content=3D\"MSHTML 6.00.2719.2200\" name=3DGENERATOR>\n",
      "<STYLE></STYLE>\n",
      "</HEAD>\n",
      "<BODY bgColor=3D#ffffff>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>Hi,</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>I have created two databas=\n",
      "es,=20\n",
      "db1&nbsp;and db2 &nbsp;in my POSTGRESQL database system.</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>And both the databases con=\n",
      "tains few=20\n",
      "tables also.</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>Could you please help me t=\n",
      "o write a=20\n",
      "query which should retrieve data from both the databases</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>(Database joins and table=\n",
      "=20\n",
      "joins)</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>Expecting and earliest=20\n",
      "reply</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>Regards</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2>Jay</FONT></DIV>\n",
      "<DIV><FONT face=3DArial color=3D#800080 size=3D2></FONT>&nbsp;</DIV></BODY>=\n",
      "</HTML>\n",
      "\n",
      "------=_NextPart_000_004E_01C2580D.1BC61260--\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(FP_train[0], FP_train[10], FP_train[20])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "From rathcairn@eircom.net  Mon Jun 24 17:05:37 2002\n",
      "Return-Path: rathcairn@eircom.net\n",
      "Delivery-Date: Tue May 21 16:22:21 2002\n",
      "Received: from mandark.labs.netnoteinc.com ([213.105.180.140]) by\n",
      "    dogma.slashnull.org (8.11.6/8.11.6) with ESMTP id g4LFARe13905 for\n",
      "    <jm@jmason.org>; Tue, 21 May 2002 16:10:29 +0100\n",
      "Received: from mail03.svc.cra.dublin.eircom.net\n",
      "    (mail03.svc.cra.dublin.eircom.net [159.134.118.19]) by\n",
      "    mandark.labs.netnoteinc.com (8.11.2/8.11.2) with SMTP id g4LFA7D26547 for\n",
      "    <jm@netnoteinc.com>; Tue, 21 May 2002 16:10:12 +0100\n",
      "Received: (qmail 26210 messnum 563502 invoked from\n",
      "    network[159.134.184.93/p93.as2.virginia1.eircom.net]); 21 May 2002\n",
      "    15:09:42 -0000\n",
      "Received: from p93.as2.virginia1.eircom.net (HELO r60qn) (159.134.184.93)\n",
      "    by mail03.svc.cra.dublin.eircom.net (qp 26210) with SMTP; 21 May 2002\n",
      "    15:09:42 -0000\n",
      "Message-Id: <00c701c200d9$a14bd540$5db8869f@r60qn>\n",
      "From: \"rathcairn\" <rathcairn@eircom.net>\n",
      "To: \"Zofia\"\n",
      "    <m22527@24h.co.jp>, <zen5@moroccomail.com>, <xxx2@ewasher.org>,\n",
      "    <x8350w24@surftimemag.com>, <wtilson19348@joymail.com>, <wfh5@wfh.org>,\n",
      "    \"Werner Gebert\"\n",
      "    <Werner.Gebert@t-online.de>,\n",
      "    \"Wendy O'Sullivan\"\n",
      "    <dwrlc@indigo.ie>,\n",
      "    \"Weedman\"\n",
      "    <fddgl@ella.hu>,\n",
      "    \"Weedman\"\n",
      "    <bxabl@dit.upm.es>,\n",
      "    \"Weedman\"\n",
      "    <uzicz@eagle.acadiau.ca>, <webmaster@korcin.com>, <web-cr@wallon.net>,\n",
      "    \"Wasted\"\n",
      "    <abykwlldfsur@egertp.ericsson.se>, <vv0033777@yahoo.com>,\n",
      "    <voiyt786@priestmail.com>, <viagraonline_376@koreainsights.com>,\n",
      "    \"Viagra to your door\"\n",
      "    <sales@globalscoop.com>, <veling@indigo.ie>,\n",
      "    \"Vanessa Roberts\"\n",
      "    <abruka@hotmail.com>,\n",
      "    \"val nelson\"\n",
      "    <valnelson@eircom.net>, <Vacation@msn.com>, <v522584@asia-links.com>,\n",
      "    <uweudj@yahoo.com>, <usave9507@eudoramail.com>, <usave783@arabia.com>,\n",
      "    \"U.S. Bachelors\"\n",
      "    <bachclub@usa.com>,\n",
      "    \"trim visitor centre\"\n",
      "    <trimvisitorcenter@eircom.net>, =?iso-8859-1?Q?Treasa_U=ED_Mh=E1irt=EDn?=\n",
      "    <treasauimhairtin@ireland.com>, <Tracy8753l27@yahoo.com>,\n",
      "    =?iso-8859-1?Q?Tracey_N=ED_Bhreathnach?= <treacybwalsh@hotmail.com>,\n",
      "    <tophatpub@tiscali.co.uk>, <tonylyons2744@spartan.dnet.co.uk>,\n",
      "    \"Tony Parker, BBC\"\n",
      "    <tony@reallife.co.uk>,\n",
      "    \"Tony McManus\"\n",
      "    <_mcmanust@eircom.net>,\n",
      "    \"Tony Corley\"\n",
      "    <t.corley@heythrop.ac.uk>, =?iso-8859-1?Q?Tom=E1s_Jim_Mac_Gearailt?=\n",
      "    <cfcdteo@iol.ie>, <tom.richardson@ie.pwcglobal.com>,\n",
      "    \"Tom O'Bryan\"\n",
      "    <obryanco@iol.ie>,\n",
      "    \"Tom Farrell\"\n",
      "    <tom@frontend.com>, <Tilliwin@aol.com>,\n",
      "    \"Terence Corish\"\n",
      "    <tc@clubi.ie>,\n",
      "    \"Tena Duarte\"\n",
      "    <drwatson8914@freemail.hu>,\n",
      "    \"Ted Creighton\"\n",
      "    <creighton@idirect.com>,\n",
      "    \"Tamas Pelyhe\"\n",
      "    <mhe@matawnet.hu>, =?iso-8859-1?Q?Tadhg_O_D=FAshl=E1ine?=\n",
      "    <Nua.Ghaeilge@may.ie>, <t3219@k.ro>, <t20476@i-france.com>,\n",
      "    <systemax20022002@yahoo.com>,\n",
      "    \"System Administrator\"\n",
      "    <postmaster@udaras.ie>,\n",
      "    \"Sunday Times\"\n",
      "    <paul.colgan@sunday-times.o.uk>,\n",
      "    \"Sunday Business Post\"\n",
      "    <sbpost@iol.ie>, <sue@hotmail.com>,\n",
      "    \"Sue Rocks\"\n",
      "    <sue@haemophilia.org.uk>,\n",
      "    \"=?iso-8859-1?B?U3Rpb2bhbiDTIENvbG3haW4=?=\"\n",
      "    <scoleman@may.ie>,\n",
      "    \"Stephen Shannon\"\n",
      "    <sshannon@meathcoco.ie>,\n",
      "    \"Stephen Price\"\n",
      "    <sprice@todayfm.com>,\n",
      "    \"Staff\"\n",
      "    <loretocrumlin.ias@tinet.ie>, <spailpin@eircom.net>,\n",
      "    <sophie.lacoste@only-yellow-pages.com>,\n",
      "    \"Smart Investor\"\n",
      "    <newsales@fl2ny.com>, <slangermanazk1@hotmail.com>, <skls@yahoo.com>,\n",
      "    <Siobhan.NiChofaigh@infineon.com>, =?iso-8859-1?Q?Siobh=E1n_N=ED_Laoire?=\n",
      "    <slaoire@celt.dias.ie>, <SIMON.BARNES@YOUR-MOVE.CO.UK>,\n",
      "    =?iso-8859-1?B?U2ltb24g0yBDcvNpbu1u?= <simoncronin@hotmail.com>,\n",
      "    =?iso-8859-1?Q?Simon_O_Cr=F3in=EDn?= <croninsimon@ireland.com>,\n",
      "    =?iso-8859-1?Q?Simon_O_Cr=F3in=EDn?= <croininsimon@ireland.com>,\n",
      "    \"Shaw Graham\"\n",
      "    <shawgraham@nireland.com>,\n",
      "    \"Shaquana Samples\"\n",
      "    <domenic.philipp@freemail.hu>,\n",
      "    \"Shane Diffily\"\n",
      "    <sdiffily@comhdhail.ie>,\n",
      "    \"SeXieRachel\"\n",
      "    <sexyrachel@braintaco.com>, =?iso-8859-1?Q?Seosaimh=EDn_N=ED_Bheaglaoich?=\n",
      "    <nibheas@rte.ie>, <sender5736x53@aol.com>, <seftopbox@stbmail.com>,\n",
      "    =?iso-8859-1?Q?Sean-N=F3s_Cois_Life?= <seannoscoislife@eircom.net>,\n",
      "    \"sean simpson\"\n",
      "    <simpsonsp@hotmail.com>,\n",
      "    \"Sean O'Sullivan\"\n",
      "    <saosull@indigo.ie>, =?iso-8859-1?Q?Se=E1n_=D3_Riain?=\n",
      "    <seanoriain@tinet.ie>, =?iso-8859-1?Q?Se=E1n_=D3_Riain?=\n",
      "    <seanoriain@eircom.net>, =?iso-8859-1?Q?Sean_=D3_Daimh=EDn?=\n",
      "    <seanodaimh@aol.com>,\n",
      "    \"Sean Mac Aoire\"\n",
      "    <seanmacaoire@netscape.net>,\n",
      "    \"Sean A O'Sullivan\"\n",
      "    <sean_a-o'sullivan@bourns.com>,\n",
      "    \"=?iso-8859-1?B?U+lhbXVzINMgTulpbGw=?=\"\n",
      "    <seamusoneill@weandus.com>,\n",
      "    \"=?iso-8859-1?B?U+lhbXVzINMgQ3Jvc+Fpbg==?=\"\n",
      "    <crosson@clubi.ie>, =?iso-8859-1?Q?S=E9amus_=D3_Connaill?=\n",
      "    <jimconnell@eircom.net>, <scooney@emirates.net.ae>, <save5749@yahoo.com>,\n",
      "    <save5459@arabia.com>,\n",
      "    \"Sandra Nic Iomhair\"\n",
      "    <antseanbheairic@eircom.net>, <sales@assuredtraffic.com>,\n",
      "    <sadataycow9345@eudoramail.com>,\n",
      "    \"sabhailteacht bothar\"\n",
      "    <sabhail@yahoo.com>, <rusheicmomra@hotmail.com>,\n",
      "    \"Ruairi O hUiginn\"\n",
      "    <ruairi@celt.dias.ie>,\n",
      "    \"Ross Robinson\"\n",
      "    <ross@wfh.org>,\n",
      "    \"Ross Carroll\"\n",
      "    <rcarroll@operation.dublin.iona.ie>,\n",
      "    \"Roselie V. Zabala\"\n",
      "    <rzabala@ns.gov.gu>, =?iso-8859-1?B?UvNpc+1uIE7tIE1oaWFu4Wlu?=\n",
      "    <rnimh@indigo.ie>, <rodney@126.com>,\n",
      "    \"Rod Allen and Barbara Quigley\"\n",
      "    <allquig@nb.sympatico.ca>,\n",
      "    \"Robbyn\"\n",
      "    <M18439@communityconnect.com>, <ro3456you@email.com>,\n",
      "    \"Richard Oakley\"\n",
      "    <oakleyrichard@hotmail.com>,\n",
      "    \"Renee Paper\"\n",
      "    <hfnv@earthlink.net>,\n",
      "    \"Rene McRogers\"\n",
      "    <Rene.McRogers.B@bayer.com>, <REJ@MoreHGH.com>,\n",
      "    \"Reggie O Riain\"\n",
      "    <oreggie@hotmail.com>,\n",
      "    \"=?iso-8859-1?B?UulhbW9ubiDTIENpYXLhaW4=?=\"\n",
      "    <reamonn.gael-linn@virgin.net>,\n",
      "    \"=?iso-8859-1?B?UulhbW9ubiDTIENpYXLhaW4=?=\"\n",
      "    <reamonn@gael-linn.net>,\n",
      "    \"Raymond Kieran\"\n",
      "    <kieran31@lineone.net>, <rathcairn@eircom.net>,\n",
      "    \"rastelli\"\n",
      "    <zq8oc1d0lp2owu@intel.com>, =?iso-8859-1?Q?Raidi=F3_na_Life?=\n",
      "    <rnl106@iol.ie>,\n",
      "    \"rafco\"\n",
      "    <rafco@maptel.es>, <r6hw45esr7actionjo@aol.com>,\n",
      "    <r1207f00@altavista.co.kr>, <prwiiw@imailbox.com>,\n",
      "    <protection_now2002@yahoo.com>, <proinsias@arainnmhor.com>,\n",
      "    \"Prescription\"\n",
      "    <ytbzb@chat.ru>, <precision_ppp@yahoo.com>, <ppl2448_01@hotmail.com>,\n",
      "    =?iso-8859-1?Q?P=F3il=EDn_Roycroft?= <schullns.ias@eircom.net>,\n",
      "    <PogCD@aol.com>, =?iso-8859-1?Q?Phyllis_U=ED_Mheadhra?=\n",
      "    <phyllisomeara@eircom.net>,\n",
      "    \"Philomena Lafferty\"\n",
      "    <philafferty@eircom.net>, <Peter.Feinsinger@NAU.EDU>,\n",
      "    \"Pekka Heikkila\"\n",
      "    <pekka.heikkila@metos.com>, =?iso-8859-1?Q?Peigi_N=ED_Chonghaile?=\n",
      "    <plearaca@tinet.ie>,\n",
      "    \"Paul Mc  Evoy\"\n",
      "    <pjdent@indigo.ie>,\n",
      "    \"Patti Kerr\"\n",
      "    <pakerr@hotmail.com>,\n",
      "    \"Patrick O'donoghue\"\n",
      "    <pod2@indigo.ie>, <patgryan@oceanfree.net>,\n",
      "    \"Pat Kavanagh\"\n",
      "    <pk@gbgroup.ie>,\n",
      "    \"Pat &Carmel Downey\"\n",
      "    <pcadowney@eircom.net>,\n",
      "    \"Parttraporn Israngkuranaayudhaya\"\n",
      "    <rapis@mucc.mahidol.ac.th>,\n",
      "    \"Pamela Aldrich\"\n",
      "    <paldrich@esatclear.ie>, <padraig.ocuinneagain2@mail.dcu.ie>,\n",
      "    =?iso-8859-1?Q?P=E1draig_Swinbourne?= <PavdiealthePub@hotmail.com>,\n",
      "    =?iso-8859-1?Q?Padraig_=D3_Cinn=E9ide?= <padraig@evolution.org>,\n",
      "    \"Padraig Nallen\"\n",
      "    <pnallen@operation.dublin.iona.ie>,\n",
      "    \"Padraig Nallen\"\n",
      "    <pnallen@operation.dublin.iona.ie>,\n",
      "    =?iso-8859-1?Q?P=E1draig_Leo_O_Curraoin?= <patrickleocurran@eircom.net>,\n",
      "    \"Paddy Clancy, The Sun\"\n",
      "    <paddyclancy@the-sun.co.uk>, <p_mcevoy@hotmail.com>, <orlag@eircom.net>,\n",
      "    <opt-in@randbad.com>, <opportunity_nox@email.com>,\n",
      "    \"Olga Vanherle\"\n",
      "    <olga@wol.be>,\n",
      "    \"Oideas Gael\"\n",
      "    <oidsgael@iol.ie>, <ocsc@ocsc.ie>,\n",
      "    \"obrienp\"\n",
      "    <obrienp@iol.ie>, <o7550k46@norikomail.com>,\n",
      "    \"North Dublin\"\n",
      "    <ndnsp@iol.ie>,\n",
      "    \"Norman Kidd\"\n",
      "    <norman.k@ukonline.co.uk>,\n",
      "    \"Noirin Nuadhain\"\n",
      "    <noirin.nuadhain@ncca.ie>,\n",
      "    \"Nils van Hinsberg\"\n",
      "    <N.P.vanHinsberg@lr.tudelft.nl>, =?iso-8859-1?Q?Nicola_N=ED_Cheallaigh?=\n",
      "    <nicolakelly@mai.ie>,\n",
      "    \"Nicky Petite\"\n",
      "    <54t7854342@msn.com>, =?iso-8859-1?Q?NIC_MHEANMAN=2C_M=C1IRE?=\n",
      "    <NICMHEAN@educ.irlgov.ie>,\n",
      "    \"Niamh McEntee\"\n",
      "    <niamh.mcentee@bookham.com>,\n",
      "    \"NI NUADHAIN, NOIRIN\"\n",
      "    <NUADHAIN@educ.irlgov.ie>, <newmanj@iobox.fi>,\n",
      "    \"Natalia Barry\"\n",
      "    <natalia@haemophilia.org.uk>,\n",
      "    \"Nadene\"\n",
      "    <lywou@24horas.com>, <n0544k76@mailcity.lycos.co.jp>,\n",
      "    =?iso-8859-1?Q?Muireann_N=ED_Fhaotain?= <marian.whelan@hotmail.com>,\n",
      "    =?iso-8859-1?Q?Muireann_N=ED_Ch=E1rthaigh?= <muireann@comhdhail.ie>,\n",
      "    <MStenhouse@mlaw.ie>,\n",
      "    \"Mr. Natural\"\n",
      "    <fddgl@comptel.fi>,\n",
      "    \"Morgan, Bernie\"\n",
      "    <morber@dunwoody.tec.mn.us>, <monicareilly@P0PMAILER.ORG>,\n",
      "    \"Moneystorm\"\n",
      "    <Professional_Series@freenet.de>,\n",
      "    \"Money in the Mail\"\n",
      "    <singledolar@hotmail.com>,\n",
      "    \"Monet\"\n",
      "    <l8075@virgilio.it>,\n",
      "    \"Moire Leyden\"\n",
      "    <moira@asti.ie>, <moiral@asti.ie>, <mireillestryjewski3884@altavista.com>,\n",
      "    <mike3561@yahoo.com>, <MicOfiach@netscape.net>,\n",
      "    \"Mick Power\"\n",
      "    <Mick@vendor.ie>, =?iso-8859-1?Q?Miche=E1l_=D3_Broin?= <obroin@iol.ie>,\n",
      "    \"Micheal O Broin\"\n",
      "    <obroin@triskel.ie>,\n",
      "    \"=?iso-8859-1?B?Te1jaGXhbCDTIEJyZWFzbOFpbg==?=\"\n",
      "    <dunloy@nacn.org>, <MichaelCroc@aol.com>,\n",
      "    \"Michael Martin\"\n",
      "    <info@reiki.ie>,\n",
      "    \"Michael & Enta Butler\"\n",
      "    <dabu.ias@eircom.net>,\n",
      "    \"mfg teo\"\n",
      "    <mfggaillimh@tinet.ie>, <merchantsworld2001@juno.com>,\n",
      "    <members@list.adult-newsletter.com>,\n",
      "    \"Megan Biddles\"\n",
      "    <8734ryhje@msn.com>,\n",
      "    \"meas media\"\n",
      "    <measmedia@eircom.net>, <mcivorpd@indigo.ie>,\n",
      "    \"Maureen Monica Griffin\"\n",
      "    <mamgriff@indiana.edu>, <MaryMarie2002av@about.com>,\n",
      "    <maryloulynne@arabia.com>,\n",
      "    \"Mary O'Shaughnessy\"\n",
      "    <mary.oshaughnessy@ucc.ie>,\n",
      "    \"Mary Ng Chua\"\n",
      "    <mary.chua@ibm.net>,\n",
      "    \"Martin O'Brien\"\n",
      "    <MartinOB@compuserve.com>,\n",
      "    \"Martin Byrne\"\n",
      "    <friction@eircom.net>, <marrisa691@cranfield.ac.uk>,\n",
      "    \"Marrinan, Shonagh \\(CAP, GCF\\)\"\n",
      "    <Shonagh.Marrinan@gecapital.com>, <mark_edwards1952@financier.com>,\n",
      "    \"Mark Eweka\"\n",
      "    <business_centre2000@yahoo.com>,\n",
      "    \"Mark\"\n",
      "    <errand1636@yahoo.com>,\n",
      "    \"Marie Gillman\"\n",
      "    <gillmam@labs.wyeth.com>,\n",
      "    \"maria hickey\"\n",
      "    <findmariahickey@hotmail.com>, <MargaretDunne@novo-wfh2000.com>,\n",
      "    <margaretculliton@unison.ie>,\n",
      "    \"Margaret Kinder\"\n",
      "    <mmk@britishlibrary.net>,\n",
      "    \"Margaret Farrington\"\n",
      "    <M.Farrington@liverpool.ac.uk>, =?iso-8859-1?Q?Mait_=D3_Br=E1daigh?=\n",
      "    <maitobradaigh@tinet.ie>, =?iso-8859-1?Q?M=E1ir=EDn_Seoighe?=\n",
      "    <dobhar@eircom.net>, =?iso-8859-1?Q?Mair=E9ad_Mic_Sh=EDom=F3in?=\n",
      "    <sidhebhean@eircom.net>, =?iso-8859-1?Q?M=E1ire_U=CD_Fhaogh=E1in?=\n",
      "    <ofaoghain@eircom.net>, =?iso-8859-1?Q?M=E1ire_U=ED_Dhufaigh?=\n",
      "    <muidhufaigh@hotmail.com>,\n",
      "    \"Maire Ni Mhainnin\"\n",
      "    <m.nimhainin@udaras.ie>, =?iso-8859-1?Q?M=E1ire_N=ED_Ghiollabh=E1in?=\n",
      "    <Maura.Kilbane@may.ie>, =?iso-8859-1?Q?M=E1ire_N=ED_Flatharta?=\n",
      "    <maire.niflatharta@daff.irlgov.ie>,\n",
      "    \"Maire Hearty\"\n",
      "    <m.hearty@udaras.ie>,\n",
      "    \"Mail Delivery System\"\n",
      "    <Mailer-Daemon@tinet.ie>, =?iso-8859-1?Q?MacTh=F3mais=2C_Uinseann?=\n",
      "    <umacthomais@ealga.ie>, <m418327@gopinoy.com>, <m1015a33@freeserve.net>,\n",
      "    <m.oreilly@unison.ie>,\n",
      "    \"Lynda Kathleen Martina Iola Costelloe\"\n",
      "    <lynda@gcom.ie>,\n",
      "    \"Louise Carroll\"\n",
      "    <jcc@nznet.gen.nz>,\n",
      "    \"Lone Pedersen\"\n",
      "    <lone@bloderforeningen.dk>, <loans_help88@yahoo.com>,\n",
      "    \"Loan Admin\"\n",
      "    <marketpro7845@gironet.nl>,\n",
      "    \"Litigation Experts\"\n",
      "    <LegalService@nueworld.com>, <line@wfh.org>, <linda.ni.ghriofa@tg4.ie>,\n",
      "    \"Lin\"\n",
      "    <bvkql@email.si>, <lifequote1aa552718@altavista.se>,\n",
      "    <life_insuranceman@yahoo.com>,\n",
      "    \"Life Insurance Quote\"\n",
      "    <fyvdv@i-france.com>, =?iso-8859-1?Q?Liam_Mac_R=E9amoinn?=\n",
      "    <murrisk@oceanfree.net>,\n",
      "    \"Liam Mac Coil\"\n",
      "    <lmaccoil@yahoo.com>,\n",
      "    \"Lee Ryan\"\n",
      "    <designroom@clubi.ie>,\n",
      "    \"Lean Doody\"\n",
      "    <Lean.Doody@operation.dublin.iona.ie>,\n",
      "    \"=?iso-8859-1?B?TOE=?=\"\n",
      "    <eolas@nuacht.com>, <l715240@zmail.to>, <kuynjy@liveclothing.com.au>,\n",
      "    <kgillin2386@mindwerk.de>,\n",
      "    \"kevin Hurley\"\n",
      "    <jkevinhurley@eircom.net>,\n",
      "    \"Kerry Woods\"\n",
      "    <kerry@tech.boo.com>,\n",
      "    \"Kellys Dunkirk\"\n",
      "    <info@kellysdunkirk.com>,\n",
      "    \"=?iso-8859-1?Q?Karen_N=ED_Bhreasl=E1in.?=\"\n",
      "    <kazbeeioi@hotmail.com>,\n",
      "    \"Karen Canty\"\n",
      "    <kcanty@eircom.net>, <Kaaren2238y32@surflondon.co.uk>,\n",
      "    \"jyogBiff Armstrong\"\n",
      "    <eoxmarmstrong6dg325sgf2@msn.com>, <jwlocke@hotmail.com>,\n",
      "    \"Justin Mason\"\n",
      "    <jm5@netnoteinc.com>,\n",
      "    \"Justin Mason\"\n",
      "    <jm@netnoteinc.com>,\n",
      "    \"Juliet Fahy \\(Work\\)\"\n",
      "    <mpbe@stjames.ie>,\n",
      "    \"Juliet Fahy \\(Home\\)\"\n",
      "    <mayday@indigo.e>,\n",
      "    \"Juliet & Paddy Fahy\"\n",
      "    <mayday@indigo.ie>,\n",
      "    \"Judy Schaefer\"\n",
      "    <jschaefer@mindspring.com>,\n",
      "    \"Judy Johnson\"\n",
      "    <sgwharton@cyberpopmail.com>,\n",
      "    \"Judy\"\n",
      "    <jfnnbptavegjzmhqok@fr.maus.de>,\n",
      "    \"Jonell Baggett\"\n",
      "    <fairs@polbox.com>, <john6518e28@hotmail.com>, <john@kerryseye.com>,\n",
      "    \"John O'Connell\"\n",
      "    <John.OConnell@may.ie>,\n",
      "    \"John Lynch\"\n",
      "    <mentrim@eircom.net>,\n",
      "    \"John Dilly\"\n",
      "    <SEPT31932@aol.com>,\n",
      "    \"John Deely\"\n",
      "    <j-deely@hoffman.cc.sophia.ac.jp>,\n",
      "    \"Joe Weiss\"\n",
      "    <joe.weiss@gmx.at>,\n",
      "    \"Joe Humphries, Irish Times\"\n",
      "    <jhumphries@irish-times.ie>,\n",
      "    \"Joe & Maureen Guilfoyle\"\n",
      "    <destiny@eircom.net>,\n",
      "    \"Jessica\"\n",
      "    <saucyjessica2000@yahoo.com>,\n",
      "    \"Jessica\"\n",
      "    <naughty_jess60691@yahoo.com>,\n",
      "    \"jenny o halloran\"\n",
      "    <princess4537@hotmail.com>,\n",
      "    \"Jenny\"\n",
      "    <jenny69@usa.net>,\n",
      "    \"Jennifer Nic an Bhaird\"\n",
      "    <jennifer.m.ward@may.ie>,\n",
      "    \"jenks\"\n",
      "    <zli31ma@usa.net>,\n",
      "    \"Javier Carrera\"\n",
      "    <ashegui@nexo.es>, <JasmineJones@msn.com>, <jamie002@arabia.com>,\n",
      "    <jamescarragher2@eircom.net>, <jagrad2@aol.com>,\n",
      "    <jacquettas@altavista.com>,\n",
      "    \"Jack O'Brien\"\n",
      "    <kjob1@mediaone.net>, <j0811t23@singmail.com>, <J.OBrien@itcilo.it>,\n",
      "    \"Ivor Ferris\"\n",
      "    <gaelach@iol.ie>, <itwill7@hotmail.com>,\n",
      "    \"Irish Haemophilia Society\"\n",
      "    <haemophiliasocitey@eircom.net>,\n",
      "    \"Irish Haemophilia Society\"\n",
      "    <haemophiliasociety@eircom.net>,\n",
      "    \"Irina Malenko\"\n",
      "    <irina.malenko@the-gem.com>,\n",
      "    \"Irina Malenko\"\n",
      "    <pravda@utvinternet.com>, <info@destressu.com>,\n",
      "    \"Info Staff\"\n",
      "    <kalil@klyvold.net>,\n",
      "    \"Info Center\"\n",
      "    <cannes211@freemail.hu>, <Imagesoft03@extra.hu>, <ilwbfiwevbs@yahoo.com>,\n",
      "    <abqewvbgf@iinet.net.au>, <aa1133777@yahoo.com>,\n",
      "    <a_crothers@looksmart.com.au>, <89ok7yumhjn@msn.com>,\n",
      "    <12unidiploma@msn.com>, <0gbvll2jf@solvo.spb.su>, <09886w@ams.com.br>,\n",
      "    <0815@activate.de>, <01bb91b2.7e887960@genesys.pt>, <00luke@upline.se>,\n",
      "    <0024@simba.nu>,\n",
      "    \"=?iso-8859-1?B?sAw=?=\"\n",
      "    <dis200@indigo.ie>\n",
      "Subject: =?iso-8859-1?Q?Fw:_CD_Nua_do_dhamhsa=ED_Ch=E9il=ED?=\n",
      "Date: Tue, 21 May 2002 16:08:40 +0100\n",
      "MIME-Version: 1.0\n",
      "X-Priority: 3\n",
      "X-Msmail-Priority: Normal\n",
      "X-Mailer: Microsoft Outlook Express 5.00.2919.6600\n",
      "Disposition-Notification-To: \"rathcairn\" <rathcairn@eircom.net>\n",
      "X-Mimeole: Produced By Microsoft MimeOLE V5.00.2919.6600\n",
      "X-Keywords: \n",
      "Content-Type: multipart/alternative; boundary=\"----=_NextPart_000_00AF_01C200E1.C5BDFF40\"\n",
      "X-ExmhMDN: ignored\n",
      "\n",
      "This is a multi-part message in MIME format.\n",
      "\n",
      "------=_NextPart_000_00AF_01C200E1.C5BDFF40\n",
      "Content-Type: text/plain;\n",
      "\tcharset=\"iso-8859-1\"\n",
      "Content-Transfer-Encoding: quoted-printable\n",
      "\n",
      "\n",
      "----- Original Message -----=20\n",
      "From: rathcairn=20\n",
      "To: Automail Script ; Automail Script ; astraea@iol.ie ; =C5se Tobin ; =\n",
      "araben16313168@hotmail.com ; aofarachain@tinet.ie ; =\n",
      "aofarachain@eircom.net ; Anthony Gorman ; Ann-Mari Bergkvist ; Annemarie =\n",
      "Meehan ; annemarie mcdonnell ; anne7172@arabia.com ; Anna Swelund ; Anna =\n",
      "Ni Thuama ; ann faherty ; anewloan@mail.ru ; Andy Plunkett ; Andrew =\n",
      "Delany ; Andrea Nic Oscair ; An Fhiontarlann ; amurtagh@tinet.ie ; =\n",
      "Amanda N=ED Ghuidhir ; als2@hotmail.com ; alpha@fun.21cn.com ; Allen =\n",
      "Moira ; Allen Carberry ; alex_doyle@ie.ibm.com ; Alan McGowan ; Alain =\n",
      "Civel ; Aishling N=ED Raghallaigh ; =C1ine M=E1ire N=ED Choile=E1in ; =\n",
      "AINE GUILFOYLE ; ailin.nichuir@ucd.ie ; Ail=EDn N=ED h=D3g=E1in ; Aileen =\n",
      "O'Meara, RTE ; advisor@physiciansblend.net ; adshwe@bekkers.com.au ; =\n",
      "adrian28@esatclear.ie ; Admin ; Ade Kallas ; adare productions ; =\n",
      "abqewvbgf@iinet.net.au ; aa1133777@yahoo.com ; =\n",
      "a_crothers@looksmart.com.au ; 89ok7yumhjn@msn.com ; 12unidiploma@msn.com =\n",
      "; 0gbvll2jf@solvo.spb.su ; Daithi Mac Carthaigh ; d7596@go.ru ; =\n",
      "d23463@portugalmail.com=20\n",
      "Cc: 09886w@ams.com.br ; 0815@activate.de ; 01bb91b2.7e887960@genesys.pt =\n",
      "; 00luke@upline.se=20\n",
      "Sent: Tuesday, May 21, 2002 3:47 PM\n",
      "Subject: CD Nua do dhamhsa=ED Ch=E9il=ED\n",
      "\n",
      "\n",
      "A chara,\n",
      "\n",
      "Email gairid le cur in i=FAl duit faoi Dhl=FAthdhiosca Nua at=E1 ar =\n",
      "f=E1il dona damhsa=ED Ch=E9il=ED is coitianta i measc daoine =F3ga.\n",
      "=20\n",
      "T=E1 Ceol ar dona Damhsa=ED seo a leanas :\n",
      "\n",
      "Balla=ED Luimn=ED\n",
      "Briseadh na Carraige\n",
      "Baint an Fh=E9ir\n",
      "Ionsa=ED na hInse\n",
      "Port an Fh=F3mhair\n",
      "Cor na S=EDog\n",
      "R=E9ic=ED Mh=E1la\n",
      "Tonna=ED Thora=ED\n",
      "Droichead =C1tha Luain\n",
      "Staic=EDn Eorna\n",
      "Se=E1in=EDn\n",
      "Cor Beirte\n",
      "Shoe the Donkey\n",
      "Waltzes=20\n",
      "agus eile\n",
      "\n",
      "Is f=E9idir an CD a cheannach tr=EDd www.damhsa.com\n",
      "n=F3 086 8339082\n",
      "\n",
      "Le meas,\n",
      "\n",
      "\n",
      "Brian =D3 Broin\n",
      "\n",
      "\n",
      "------=_NextPart_000_00AF_01C200E1.C5BDFF40\n",
      "Content-Type: text/html;\n",
      "\tcharset=\"iso-8859-1\"\n",
      "Content-Transfer-Encoding: quoted-printable\n",
      "\n",
      "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML 4.0 Transitional//EN\">\n",
      "<HTML><HEAD>\n",
      "<META content=3D\"text/html; charset=3Diso-8859-1\" =\n",
      "http-equiv=3DContent-Type>\n",
      "<META content=3D\"MSHTML 5.00.2919.6307\" name=3DGENERATOR>\n",
      "<STYLE></STYLE>\n",
      "</HEAD>\n",
      "<BODY bgColor=3D#ffffff>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV style=3D\"FONT: 10pt arial\">----- Original Message -----=20\n",
      "<DIV style=3D\"BACKGROUND: #e4e4e4; font-color: black\"><B>From:</B> <A=20\n",
      "href=3D\"mailto:rathcairn@eircom.net\" =\n",
      "title=3Drathcairn@eircom.net>rathcairn</A>=20\n",
      "</DIV>\n",
      "<DIV><B>To:</B> <A href=3D\"mailto:nobody@tinet.ie\" =\n",
      "title=3Dnobody@tinet.ie>Automail=20\n",
      "Script</A> ; <A href=3D\"mailto:nobody@eircom.net\" =\n",
      "title=3Dnobody@eircom.net>Automail=20\n",
      "Script</A> ; <A href=3D\"mailto:astraea@iol.ie\"=20\n",
      "title=3Dastraea@iol.ie>astraea@iol.ie</A> ; <A =\n",
      "href=3D\"mailto:aasvtobi@online.no\"=20\n",
      "title=3Daasvtobi@online.no>=C5se Tobin</A> ; <A=20\n",
      "href=3D\"mailto:araben16313168@hotmail.com\"=20\n",
      "title=3Daraben16313168@hotmail.com>araben16313168@hotmail.com</A> ; <A=20\n",
      "href=3D\"mailto:aofarachain@tinet.ie\"=20\n",
      "title=3Daofarachain@tinet.ie>aofarachain@tinet.ie</A> ; <A=20\n",
      "href=3D\"mailto:aofarachain@eircom.net\"=20\n",
      "title=3Daofarachain@eircom.net>aofarachain@eircom.net</A> ; <A=20\n",
      "href=3D\"mailto:anthonygorman@eircom.net\" =\n",
      "title=3Danthonygorman@eircom.net>Anthony=20\n",
      "Gorman</A> ; <A href=3D\"mailto:Ann-Mari.Bergkvist@lh.luth.se\"=20\n",
      "title=3DAnn-Mari.Bergkvist@lh.luth.se>Ann-Mari Bergkvist</A> ; <A=20\n",
      "href=3D\"mailto:annemarie.meehan@barnardos.ie\"=20\n",
      "title=3Dannemarie.meehan@barnardos.ie>Annemarie Meehan</A> ; <A=20\n",
      "href=3D\"mailto:annemariemcdonnell@ireland.com\"=20\n",
      "title=3Dannemariemcdonnell@ireland.com>annemarie mcdonnell</A> ; <A=20\n",
      "href=3D\"mailto:anne7172@arabia.com\"=20\n",
      "title=3Danne7172@arabia.com>anne7172@arabia.com</A> ; <A=20\n",
      "href=3D\"mailto:Anna.Swelund@ericsson.com\" =\n",
      "title=3DAnna.Swelund@ericsson.com>Anna=20\n",
      "Swelund</A> ; <A href=3D\"mailto:anna@glornangael.ie\"=20\n",
      "title=3Danna@glornangael.ie>Anna Ni Thuama</A> ; <A=20\n",
      "href=3D\"mailto:annfaherty@esatclear.ie\" =\n",
      "title=3Dannfaherty@esatclear.ie>ann faherty=20\n",
      "</A>; <A href=3D\"mailto:anewloan@mail.ru\"=20\n",
      "title=3Danewloan@mail.ru>anewloan@mail.ru</A> ; <A=20\n",
      "href=3D\"mailto:Plunkett@gofree.Indigo.ie\" =\n",
      "title=3DPlunkett@gofree.Indigo.ie>Andy=20\n",
      "Plunkett</A> ; <A href=3D\"mailto:delanya@cs.tcd.ie\" =\n",
      "title=3Ddelanya@cs.tcd.ie>Andrew=20\n",
      "Delany</A> ; <A href=3D\"mailto:naiscoilardmhacha@cinni.org\"=20\n",
      "title=3Dnaiscoilardmhacha@cinni.org>Andrea Nic Oscair</A> ; <A=20\n",
      "href=3D\"mailto:fntrdan@oceanfree.net\" title=3Dfntrdan@oceanfree.net>An=20\n",
      "Fhiontarlann</A> ; <A href=3D\"mailto:amurtagh@tinet.ie\"=20\n",
      "title=3Damurtagh@tinet.ie>amurtagh@tinet.ie</A> ; <A=20\n",
      "href=3D\"mailto:amandaeile@hotmail.com\" =\n",
      "title=3Damandaeile@hotmail.com>Amanda N=ED=20\n",
      "Ghuidhir</A> ; <A href=3D\"mailto:als2@hotmail.com\"=20\n",
      "title=3Dals2@hotmail.com>als2@hotmail.com</A> ; <A=20\n",
      "href=3D\"mailto:alpha@fun.21cn.com\" =\n",
      "title=3Dalpha@fun.21cn.com>alpha@fun.21cn.com</A>=20\n",
      "; <A href=3D\"mailto:hillview@eircom.net\" =\n",
      "title=3Dhillview@eircom.net>Allen Moira</A>=20\n",
      "; <A href=3D\"mailto:acarberr@iol.ie\" title=3Dacarberr@iol.ie>Allen =\n",
      "Carberry</A> ; <A=20\n",
      "href=3D\"mailto:alex_doyle@ie.ibm.com\"=20\n",
      "title=3Dalex_doyle@ie.ibm.com>alex_doyle@ie.ibm.com</A> ; <A=20\n",
      "href=3D\"mailto:alan.mcgowan@kerstenhunik.ie\"=20\n",
      "title=3Dalan.mcgowan@kerstenhunik.ie>Alan McGowan</A> ; <A=20\n",
      "href=3D\"mailto:amcivel@club-internet.fr\" =\n",
      "title=3Damcivel@club-internet.fr>Alain=20\n",
      "Civel</A> ; <A href=3D\"mailto:aishlingreilly@hotmail.com\"=20\n",
      "title=3Daishlingreilly@hotmail.com>Aishling N=ED Raghallaigh</A> ; <A=20\n",
      "href=3D\"mailto:sccmh@iol.ie\" title=3Dsccmh@iol.ie>=C1ine M=E1ire N=ED =\n",
      "Choile=E1in</A> ; <A=20\n",
      "href=3D\"mailto:aine.guilfoyle@NUIGALWAY.IE\" =\n",
      "title=3Daine.guilfoyle@NUIGALWAY.IE>AINE=20\n",
      "GUILFOYLE</A> ; <A href=3D\"mailto:ailin.nichuir@ucd.ie\"=20\n",
      "title=3Dailin.nichuir@ucd.ie>ailin.nichuir@ucd.ie</A> ; <A=20\n",
      "href=3D\"mailto:aileenhogan@yahoo.com\" =\n",
      "title=3Daileenhogan@yahoo.com>Ail=EDn N=ED=20\n",
      "h=D3g=E1in</A> ; <A href=3D\"mailto:omearaa@rte.ie\" =\n",
      "title=3Domearaa@rte.ie>Aileen=20\n",
      "O'Meara, RTE</A> ; <A href=3D\"mailto:advisor@physiciansblend.net\"=20\n",
      "title=3Dadvisor@physiciansblend.net>advisor@physiciansblend.net</A> ; <A =\n",
      "\n",
      "href=3D\"mailto:adshwe@bekkers.com.au\"=20\n",
      "title=3Dadshwe@bekkers.com.au>adshwe@bekkers.com.au</A> ; <A=20\n",
      "href=3D\"mailto:adrian28@esatclear.ie\"=20\n",
      "title=3Dadrian28@esatclear.ie>adrian28@esatclear.ie</A> ; <A=20\n",
      "href=3D\"mailto:vdkpn@jubiipost.dk\" title=3Dvdkpn@jubiipost.dk>Admin</A> =\n",
      "; <A=20\n",
      "href=3D\"mailto:ade@cut.ee\" title=3Dade@cut.ee>Ade Kallas</A> ; <A=20\n",
      "href=3D\"mailto:adare@eircom.net\" title=3Dadare@eircom.net>adare =\n",
      "productions</A> ; <A=20\n",
      "href=3D\"mailto:abqewvbgf@iinet.net.au\"=20\n",
      "title=3Dabqewvbgf@iinet.net.au>abqewvbgf@iinet.net.au</A> ; <A=20\n",
      "href=3D\"mailto:aa1133777@yahoo.com\"=20\n",
      "title=3Daa1133777@yahoo.com>aa1133777@yahoo.com</A> ; <A=20\n",
      "href=3D\"mailto:a_crothers@looksmart.com.au\"=20\n",
      "title=3Da_crothers@looksmart.com.au>a_crothers@looksmart.com.au</A> ; <A =\n",
      "\n",
      "href=3D\"mailto:89ok7yumhjn@msn.com\"=20\n",
      "title=3D89ok7yumhjn@msn.com>89ok7yumhjn@msn.com</A> ; <A=20\n",
      "href=3D\"mailto:12unidiploma@msn.com\"=20\n",
      "title=3D12unidiploma@msn.com>12unidiploma@msn.com</A> ; <A=20\n",
      "href=3D\"mailto:0gbvll2jf@solvo.spb.su\"=20\n",
      "title=3D0gbvll2jf@solvo.spb.su>0gbvll2jf@solvo.spb.su</A> ; <A=20\n",
      "href=3D\"mailto:daithimaccarthaigh@eircom.net\"=20\n",
      "title=3Ddaithimaccarthaigh@eircom.net>Daithi Mac Carthaigh</A> ; <A=20\n",
      "href=3D\"mailto:d7596@go.ru\" title=3Dd7596@go.ru>d7596@go.ru</A> ; <A=20\n",
      "href=3D\"mailto:d23463@portugalmail.com\"=20\n",
      "title=3Dd23463@portugalmail.com>d23463@portugalmail.com</A> </DIV>\n",
      "<DIV><B>Cc:</B> <A href=3D\"mailto:09886w@ams.com.br\"=20\n",
      "title=3D09886w@ams.com.br>09886w@ams.com.br</A> ; <A=20\n",
      "href=3D\"mailto:0815@activate.de\" =\n",
      "title=3D0815@activate.de>0815@activate.de</A> ; <A=20\n",
      "href=3D\"mailto:01bb91b2.7e887960@genesys.pt\"=20\n",
      "title=3D01bb91b2.7e887960@genesys.pt>01bb91b2.7e887960@genesys.pt</A> ; =\n",
      "<A=20\n",
      "href=3D\"mailto:00luke@upline.se\" =\n",
      "title=3D00luke@upline.se>00luke@upline.se</A>=20\n",
      "</DIV>\n",
      "<DIV><B>Sent:</B> Tuesday, May 21, 2002 3:47 PM</DIV>\n",
      "<DIV><B>Subject:</B> CD Nua do dhamhsa=ED Ch=E9il=ED</DIV></DIV>\n",
      "<DIV><BR></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>A chara,</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Email gairid le cur in =\n",
      "i=FAl duit faoi=20\n",
      "Dhl=FAthdhiosca Nua at=E1 ar f=E1il dona damhsa=ED Ch=E9il=ED is =\n",
      "coitianta i measc daoine=20\n",
      "=F3ga.</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2></FONT>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>T=E1 Ceol ar dona =\n",
      "Damhsa=ED seo a leanas=20\n",
      ":</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Balla=ED =\n",
      "Luimn=ED</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Briseadh na =\n",
      "Carraige</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Baint an =\n",
      "Fh=E9ir</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Ionsa=ED na =\n",
      "hInse</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Port an =\n",
      "Fh=F3mhair</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Cor na =\n",
      "S=EDog</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>R=E9ic=ED =\n",
      "Mh=E1la</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Tonna=ED =\n",
      "Thora=ED</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Droichead =C1tha =\n",
      "Luain</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Staic=EDn =\n",
      "Eorna</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Se=E1in=EDn</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Cor Beirte</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Shoe the =\n",
      "Donkey</FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" =\n",
      "size=3D2>Waltzes</FONT>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>agus eile</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Is f=E9idir an CD a =\n",
      "cheannach tr=EDd <A=20\n",
      "href=3D\"http://www.damhsa.com\">www.damhsa.com</A></FONT></DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>n=F3 086 =\n",
      "8339082</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Le meas,</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV>&nbsp;</DIV>\n",
      "<DIV><FONT face=3D\"Bookman Old Style\" size=3D2>Brian =D3 =\n",
      "Broin</FONT></DIV>\n",
      "<DIV>&nbsp;</DIV></BODY></HTML>\n",
      "\n",
      "------=_NextPart_000_00AF_01C200E1.C5BDFF40--\n",
      "\n",
      " From social-admin@linux.ie  Tue Aug  6 12:52:55 2002\n",
      "Return-Path: <social-admin@linux.ie>\n",
      "Delivered-To: yyyy@localhost.netnoteinc.com\n",
      "Received: from localhost (localhost [127.0.0.1])\n",
      "\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id 4B1E644242\n",
      "\tfor <jm@localhost>; Tue,  6 Aug 2002 07:21:33 -0400 (EDT)\n",
      "Received: from phobos [127.0.0.1]\n",
      "\tby localhost with IMAP (fetchmail-5.9.0)\n",
      "\tfor jm@localhost (single-drop); Tue, 06 Aug 2002 12:21:33 +0100 (IST)\n",
      "Received: from lugh.tuatha.org (root@lugh.tuatha.org [194.125.145.45]) by\n",
      "    dogma.slashnull.org (8.11.6/8.11.6) with ESMTP id g744giv16318 for\n",
      "    <jm+ilug-social@jmason.org>; Sun, 4 Aug 2002 05:42:44 +0100\n",
      "Received: from lugh (root@localhost [127.0.0.1]) by lugh.tuatha.org\n",
      "    (8.9.3/8.9.3) with ESMTP id FAA29962; Sun, 4 Aug 2002 05:40:39 +0100\n",
      "Received: from mx07.routeit21.com ([64.25.34.45]) by lugh.tuatha.org\n",
      "    (8.9.3/8.9.3) with ESMTP id FAA29925 for <social@linux.ie>; Sun,\n",
      "    4 Aug 2002 05:40:33 +0100\n",
      "From: aljimenez@mx08.routeit21.com\n",
      "X-Authentication-Warning: lugh.tuatha.org: Host [64.25.34.45] claimed to\n",
      "    be mx07.routeit21.com\n",
      "Date: Sun, 4 Aug 2002 08:49:23 -0400\n",
      "Message-Id: <200208041249.g74CnNH24959@mx07.routeit21.com>\n",
      "X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.4.6-pre7-xfs-jfs i586)\n",
      "Reply-To: <bbjkkll@mx08.routeit21.com>\n",
      "To: <social@quebectel.com>\n",
      "Subject: [ILUG-Social] Mortgage Rates Are Down.  isoxtjcn\n",
      "Sender: social-admin@linux.ie\n",
      "Errors-To: social-admin@linux.ie\n",
      "X-Mailman-Version: 1.1\n",
      "Precedence: bulk\n",
      "List-Id: Irish Linux Users' Group social events <social.linux.ie>\n",
      "X-Beenthere: social@linux.ie\n",
      "\n",
      "When America's top companies compete for your business, you win.\n",
      "\n",
      "http://getit.routeit21.com/sure_quote/\n",
      "\n",
      "Take a moment to let us show you that we are here to save you money and address your concerns with absolutely no hassle, no obligation, no cost quotes, on all your needs, from America's top companies.\n",
      "\n",
      "\n",
      "-Mortgage rates that save you thousands.\n",
      "-New home loans.\n",
      "-Refinance or consolidate high interest credit card debt into a low interest mortgage.\n",
      "\n",
      "\n",
      "http://getit.routeit21.com/sure_quote/\n",
      "\n",
      "\n",
      "\"...was able to get 3 great offers in\n",
      "less than 24 hours.\" -Jennifer C\n",
      "\n",
      "\"Met all my needs... being able to search\n",
      "for loans in a way that puts me in control.\" -Robert T.\n",
      "\n",
      "\"..it was easy, effortless...!\"-Susan A.\n",
      "\n",
      " \n",
      "\n",
      "Click here to delete your address from future updates.\n",
      "http://getit.routeit21.com/sure_quote/rm/\n",
      "\n",
      "-- \n",
      "Irish Linux Users' Group Social Events: social@linux.ie\n",
      "http://www.linux.ie/mailman/listinfo/social for (un)subscription information.\n",
      "List maintainer: listmaster@linux.ie\n",
      "\n",
      "\n",
      " From fork-admin@xent.com  Tue Aug  6 12:53:51 2002\n",
      "Return-Path: <fork-admin@xent.com>\n",
      "Delivered-To: yyyy@localhost.netnoteinc.com\n",
      "Received: from localhost (localhost [127.0.0.1])\n",
      "\tby phobos.labs.netnoteinc.com (Postfix) with ESMTP id A4C6144201\n",
      "\tfor <jm@localhost>; Tue,  6 Aug 2002 07:25:02 -0400 (EDT)\n",
      "Received: from phobos [127.0.0.1]\n",
      "\tby localhost with IMAP (fetchmail-5.9.0)\n",
      "\tfor jm@localhost (single-drop); Tue, 06 Aug 2002 12:25:02 +0100 (IST)\n",
      "Received: from xent.com ([64.161.22.236]) by dogma.slashnull.org\n",
      "    (8.11.6/8.11.6) with ESMTP id g74NHsv13071 for <jm@jmason.org>;\n",
      "    Mon, 5 Aug 2002 00:17:55 +0100\n",
      "Received: from lair.xent.com (localhost [127.0.0.1]) by xent.com (Postfix)\n",
      "    with ESMTP id 3F33B2940EB; Sun,  4 Aug 2002 16:15:04 -0700 (PDT)\n",
      "Delivered-To: fork@spamassassin.taint.org\n",
      "Received: from sm14.texas.rr.com (sm14.texas.rr.com [24.93.35.41]) by\n",
      "    xent.com (Postfix) with ESMTP id 4B62D2940EA for <fork@xent.com>;\n",
      "    Sun,  4 Aug 2002 16:14:03 -0700 (PDT)\n",
      "Received: from waynemain (cs2427121-96.houston.rr.com [24.27.121.96]) by\n",
      "    sm14.texas.rr.com (8.12.0.Beta16/8.12.0.Beta16) with ESMTP id\n",
      "    g74NG4Uw031613 for <fork@xent.com>; Sun, 4 Aug 2002 18:16:10 -0500\n",
      "Message-Id: <4110-22002804231443982@waynemain>\n",
      "From: \"Wayne\" <lwharris@flash.net>\n",
      "To: \"fork@spamassassin.taint.org\" <fork@spamassassin.taint.org>\n",
      "Subject: OOPS!  Please Excuse My mistake!\n",
      "MIME-Version: 1.0\n",
      "Content-Type: text/plain; charset=US-ASCII\n",
      "Sender: fork-admin@xent.com\n",
      "Errors-To: fork-admin@xent.com\n",
      "X-Beenthere: fork@spamassassin.taint.org\n",
      "X-Mailman-Version: 2.0.11\n",
      "Precedence: bulk\n",
      "List-Help: <mailto:fork-request@xent.com?subject=help>\n",
      "List-Post: <mailto:fork@spamassassin.taint.org>\n",
      "List-Subscribe: <http://xent.com/mailman/listinfo/fork>, <mailto:fork-request@xent.com?subject=subscribe>\n",
      "List-Id: Friends of Rohit Khare <fork.xent.com>\n",
      "List-Unsubscribe: <http://xent.com/mailman/listinfo/fork>,\n",
      "    <mailto:fork-request@xent.com?subject=unsubscribe>\n",
      "List-Archive: <http://xent.com/pipermail/fork/>\n",
      "Date: Sun, 4 Aug 2002 18:14:44 -0500\n",
      "X-MIME-Autoconverted: from quoted-printable to 8bit by dogma.slashnull.org\n",
      "    id g74NHsv13071\n",
      "Content-Transfer-Encoding: 8bit\n",
      "\n",
      "Just sent you a note with the wrong link. Tthe correct one is \n",
      "\n",
      "http://letscreatebiz.com/whjoinfree  \n",
      "\n",
      "It is worth a minute of your time to check it out.\n",
      "\n",
      "\n",
      "Remember, you have nothing to lose and potentially a lot to gain!\n",
      "\n",
      "\n",
      "Best Regards,\n",
      "\n",
      "Wayne\n",
      "\n",
      "---------------------------------------------------\n",
      "To be removed from future mailings, simply\n",
      "respond with REMOVE in the subject line.\n",
      "\n",
      "\n",
      "http://xent.com/mailman/listinfo/fork\n",
      "\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(FN_train[0], FN_train[10], FN_train[20])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model_saver(model, saving_name, saving_path=MODEL_SAVING_PATH):\n",
    "    from sklearn.externals import joblib\n",
    "    os.makedirs(MODEL_SAVING_PATH, exist_ok=True)\n",
    "    joblib.dump(model, os.path.join(saving_path, saving_name))\n",
    "    print('Model saved in ', saving_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "sgd_clf_final.fit(X_train, y_train)\n",
    "model_saver(sgd_clf_final, 'sgd_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### *Random Forests*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 96 candidates, totalling 288 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "C:\\Users\\ghare\\Anaconda3\\lib\\site-packages\\sklearn\\externals\\joblib\\externals\\loky\\process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed: 43.4min\n",
      "[Parallel(n_jobs=-1)]: Done 154 tasks      | elapsed: 157.5min\n",
      "[Parallel(n_jobs=-1)]: Done 288 out of 288 | elapsed: 200.5min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('mess', MessagePunctSelector(add_main_par=True, keep_header=False, keep_punct=False,\n",
       "           keep_subject=False, mask_URL=True, mask_number=True)), ('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='...ators=500, n_jobs=None,\n",
       "            oob_score=False, random_state=42, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], 'mess__mask_number': [True, False], 'mess__keep_punct': [True, False], 'vect__lowercase': [True], 'vect__tokenizer': [<__main__.StemTokenizer object at 0x0000025883C08588>, <__main_...enizer': [None], 'vect__token_pattern': ['\\\\w\\\\w+', '\\\\w\\\\w+|[!\"#%&\\'()*+,-./:;<=>?@[\\\\]^_`{|}~]']}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_clf = Pipeline([\n",
    "    ('mess', MessagePunctSelector()),\n",
    "    ('vect', CountVectorizer()),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('rf', RandomForestClassifier(n_estimators=500, random_state=42))\n",
    "])\n",
    "param_grid = [\n",
    "    {'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], \n",
    "     'mess__mask_number': [True, False], 'mess__keep_punct': [True, False], 'vect__lowercase': [True],\n",
    "     'vect__tokenizer': [StemTokenizer(), LemmaTokenizer()]\n",
    "     },\n",
    "    {'mess__keep_header': [True, False], 'mess__keep_subject': [True, False], 'mess__mask_URL': [True, False], \n",
    "     'mess__mask_number': [True, False], 'mess__keep_punct': [True], 'vect__lowercase': [True], \n",
    "     'vect__tokenizer': [None], 'vect__token_pattern': ['\\w\\w+', '''\\w\\w+|[!\"#%&'()*+,-./:;<=>?@[\\]^_`{|}~]''']\n",
    "     } #specific custom pattern for tokenization with/wo punctuation\n",
    "]\n",
    "cv = GridSearchCV(rf_clf, param_grid=param_grid, cv=3, verbose=2, n_jobs=-1, scoring='average_precision')\n",
    "cv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.35%, best parameters : {'mess__keep_header': True, 'mess__keep_punct': True, 'mess__keep_subject': True, 'mess__mask_URL': True, 'mess__mask_number': True, 'vect__lowercase': True, 'vect__token_pattern': '\\\\w\\\\w+', 'vect__tokenizer': None}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 27 candidates, totalling 81 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed: 10.0min\n",
      "[Parallel(n_jobs=-1)]: Done  81 out of  81 | elapsed: 22.4min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('mess', MessagePunctSelector(add_main_par=True, keep_header=True, keep_punct=True,\n",
       "           keep_subject=True, mask_URL=True, mask_number=True)), ('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='con...ators=500, n_jobs=None,\n",
       "            oob_score=False, random_state=42, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'vect__ngram_range': [(1, 1), (1, 2), (2, 2)], 'vect__max_df': [1.0, 0.8, 0.6], 'vect__min_df': [1, 0.001, 0.005]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_clf = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=True, keep_punct=True, \n",
    "                                  mask_URL=True, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=None, token_pattern='\\\\w\\\\w+')),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('rf', RandomForestClassifier(n_estimators=500, random_state=42))\n",
    "])\n",
    "param_grid = [\n",
    "    {'vect__ngram_range': [(1, 1), (1, 2), (2, 2)], 'vect__max_df': [1.0, 0.8, 0.6], \n",
    "     'vect__min_df': [1, 0.001, 0.005]\n",
    "     }\n",
    "]\n",
    "cv = GridSearchCV(rf_clf, param_grid=param_grid, cv=3, verbose=2, n_jobs=-1, scoring='average_precision')\n",
    "cv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.64%, best parameters : {'vect__max_df': 0.8, 'vect__min_df': 0.005, 'vect__ngram_range': (1, 2)}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 4 candidates, totalling 12 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  12 out of  12 | elapsed:  3.1min remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done  12 out of  12 | elapsed:  3.1min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=Pipeline(memory=None,\n",
       "     steps=[('mess', MessagePunctSelector(add_main_par=True, keep_header=True, keep_punct=True,\n",
       "           keep_subject=True, mask_URL=True, mask_number=True)), ('vect', CountVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, encoding='utf-8', input='con...ators=500, n_jobs=None,\n",
       "            oob_score=False, random_state=42, verbose=0, warm_start=False))]),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'vect__ngram_range': [(1, 2)], 'vect__max_df': [0.8, 0.6], 'vect__min_df': [0.005, 0.001]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_clf = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=True, keep_punct=True, \n",
    "                                  mask_URL=True, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=None, token_pattern='\\\\w\\\\w+')),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('rf', RandomForestClassifier(n_estimators=500, random_state=42))\n",
    "])\n",
    "param_grid = [\n",
    "    {'vect__ngram_range': [(1, 2)], 'vect__max_df': [0.8, 0.6], \n",
    "     'vect__min_df': [0.005, 0.001]\n",
    "     }\n",
    "]\n",
    "cv = GridSearchCV(rf_clf, param_grid=param_grid, cv=3, verbose=2, n_jobs=-1, scoring='average_precision')\n",
    "cv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.64%, best parameters : {'vect__max_df': 0.8, 'vect__min_df': 0.005, 'vect__ngram_range': (1, 2)}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "prep_pip = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=True, keep_punct=True, \n",
    "                                  mask_URL=True, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=None, token_pattern='\\\\w\\\\w+', \n",
    "                             max_df=0.8, min_df=0.005, ngram_range=(1, 2))),\n",
    "    ('tfidf', TfidfTransformer())\n",
    "])\n",
    "\n",
    "X_prep = prep_pip.fit_transform(X_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 30 candidates, totalling 90 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  33 tasks      | elapsed:  2.0min\n",
      "[Parallel(n_jobs=-1)]: Done  90 out of  90 | elapsed:  5.5min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=3, error_score='raise-deprecating',\n",
       "       estimator=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=None,\n",
       "            oob_score=False, random_state=42, verbose=0, warm_start=False),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid={'max_features': ['auto', 'log2', 0.01], 'min_samples_split': [2, 0.05, 0.1, 0.15, 0.2], 'criterion': ['gini', 'entropy']},\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score='warn',\n",
       "       scoring='average_precision', verbose=2)"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rf_clf = RandomForestClassifier(random_state=42, n_estimators=500)\n",
    "\n",
    "\n",
    "param_grid = {'max_features': ['auto', 'log2', 0.01], \n",
    "              'min_samples_split': [2, 0.05, 0.1, 0.15, 0.2], 'criterion': ['gini', 'entropy']}\n",
    "\n",
    "cv = GridSearchCV(rf_clf, param_grid=param_grid, scoring='average_precision', cv=3, verbose=2, n_jobs=-1)\n",
    "cv.fit(X_prep, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best score : 99.64%, best parameters : {'criterion': 'gini', 'max_features': 0.01, 'min_samples_split': 2}\n"
     ]
    }
   ],
   "source": [
    "print('Best score : {:.2f}%, best parameters : {}'.format(cv.best_score_ * 100, cv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done   5 out of   5 | elapsed:  1.2min finished\n"
     ]
    }
   ],
   "source": [
    "rf_clf_final = Pipeline([\n",
    "    ('mess', MessagePunctSelector(keep_header=True, keep_subject=True, keep_punct=True, \n",
    "                                  mask_URL=True, mask_number=True)),\n",
    "    ('vect', CountVectorizer(lowercase=True, tokenizer=None, token_pattern='\\\\w\\\\w+', \n",
    "                             max_df=0.8, min_df=0.005, ngram_range=(1, 2))),\n",
    "    ('tfidf', TfidfTransformer()),\n",
    "    ('rf', RandomForestClassifier(n_estimators=500, random_state=42, max_features=0.01))\n",
    "])\n",
    "y_pred = cross_val_predict(rf_clf_final, X_train, y_train, cv=5, verbose=2, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy score on validation set: 98.01%\n",
      "Final precision score on validation set: 96.65%\n",
      "Final recall score on validation set: 97.03%\n"
     ]
    }
   ],
   "source": [
    "print(\"Final accuracy score on validation set: {:.2f}%\".format(accuracy_score(y_train, y_pred)*100))\n",
    "print(\"Final precision score on validation set: {:.2f}%\".format(precision_score(y_train, y_pred)*100))\n",
    "print(\"Final recall score on validation set: {:.2f}%\".format(recall_score(y_train, y_pred)*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model saved in  ./Models/\n"
     ]
    }
   ],
   "source": [
    "rf_clf_final.fit(X_train, y_train)\n",
    "model_saver(rf_clf_final, 'rf_clf.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### **Evaluating the model on the test set**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final accuracy score on test set: 99.34%\n",
      "Final precision score on test set: 98.69%\n",
      "Final recall score on test set: 99.21%\n"
     ]
    }
   ],
   "source": [
    "y_test_pred = sgd_clf_final.predict(X_test)\n",
    "print(\"Final accuracy score on test set: {:.2f}%\".format(accuracy_score(y_test, y_test_pred)*100))\n",
    "print(\"Final precision score on test set: {:.2f}%\".format(precision_score(y_test, y_test_pred)*100))\n",
    "print(\"Final recall score on test set: {:.2f}%\".format(recall_score(y_test, y_test_pred)*100))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}