Merge pull request #682 from flairNLP/add-asahi-shimbun

Add `AsahiShimbun`

docs/supported_publishers.md

@@ -1254,6 +1254,21 @@
     </tr>
   </thead>
   <tbody>
+    <tr>
+      <td>
+        <code>AsahiShimbun</code>
+      </td>
+      <td>
+        <div>Asahi Shimbun</div>
+      </td>
+      <td>
+        <a href="https://www.asahi.com/">
+          <span>www.asahi.com</span>
+        </a>
+      </td>
+      <td>&#160;</td>
+      <td>&#160;</td>
+    </tr>
     <tr>
       <td>
         <code>TheJapanNews</code>

src/fundus/publishers/jp/__init__.py

@@ -1,4 +1,5 @@
 from fundus.publishers.base_objects import Publisher, PublisherGroup
+from fundus.publishers.jp.asahi_shimbun import AsahiShimbunParser
 from fundus.publishers.jp.the_japan_news import TheJapanNewsParser
 from fundus.publishers.jp.yomiuri_shimbun import YomiuriShimbunParser
 from fundus.scraping.filter import regex_filter
@@ -28,3 +29,10 @@ class JP(metaclass=PublisherGroup):
             NewsMap("https://www.yomiuri.co.jp/sitemap-news-latest.xml"),
         ],
     )
+
+    AsahiShimbun = Publisher(
+        name="Asahi Shimbun",
+        domain="https://www.asahi.com/",
+        parser=AsahiShimbunParser,
+        sources=[NewsMap("https://www.asahi.com/sitemap.xml")],
+    )

src/fundus/publishers/jp/asahi_shimbun.py

@@ -0,0 +1,60 @@
+import datetime
+import re
+from typing import List, Optional
+
+from lxml.cssselect import CSSSelector
+
+from fundus.parser import ArticleBody, BaseParser, Image, ParserProxy, attribute
+from fundus.parser.utility import (
+    apply_substitution_pattern_over_list,
+    extract_article_body_with_selector,
+    generic_author_parsing,
+    generic_date_parsing,
+    generic_topic_parsing,
+    image_extraction,
+)
+
+
+class AsahiShimbunParser(ParserProxy):
+    class V1(BaseParser):
+        _summary_selector = CSSSelector("div.nfyQp > div.bv2Sj > p")
+        _paragraph_selector = CSSSelector("div.nfyQp > p")
+        _subtitle_selector = CSSSelector("div.nfyQp > h2")
+
+        topic_bloat_pattern = re.compile(r"朝日新聞デジタル|朝日新聞|ニュース|新聞|その他・話題")
+
+        @attribute
+        def body(self) -> Optional[ArticleBody]:
+            return extract_article_body_with_selector(
+                self.precomputed.doc,
+                paragraph_selector=self._paragraph_selector,
+                summary_selector=self._summary_selector,
+                subheadline_selector=self._subtitle_selector,
+            )
+
+        @attribute
+        def title(self) -> Optional[str]:
+            return self.precomputed.meta.get("TITLE")
+
+        @attribute
+        def publishing_date(self) -> Optional[datetime.datetime]:
+            return generic_date_parsing(self.precomputed.meta.get("article:published_time"))
+
+        @attribute
+        def authors(self) -> List[str]:
+            return generic_author_parsing(self.precomputed.ld.bf_search("author"))
+
+        @attribute
+        def topics(self) -> List[str]:
+            return apply_substitution_pattern_over_list(
+                generic_topic_parsing(self.precomputed.meta.get("keywords")), self.topic_bloat_pattern
+            )
+
+        @attribute
+        def images(self) -> List[Image]:
+            return image_extraction(
+                doc=self.precomputed.doc,
+                paragraph_selector=self._paragraph_selector,
+                author_selector=re.compile(r"、(?P<credits>[^、]*?)撮影"),
+                relative_urls=True,
+            )

tests/resources/parser/test_data/jp/AsahiShimbun.json

@@ -0,0 +1,52 @@
+{
+  "V1": {
+    "authors": [
+      "朝日新聞デジタル"
+    ],
+    "body": {
+      "summary": [],
+      "sections": [
+        {
+          "headline": [],
+          "paragraphs": [
+            "気象庁は13日、午後9時19分ごろ、日向灘（北緯31.8度、東経131.6度）で震度5弱の地震があったと発表した。震源の深さは約30キロ、地震の規模（マグニチュード）は6.9と推定される。この地震で、気象庁は高知県と宮崎県に1メートルの津波注意報を出した。",
+            "各地の震度は次のとおり。",
+            "＜震度5弱＞",
+            "宮崎県：高鍋町、新富町、宮崎市",
+            "＜震度4＞",
+            "宮崎県：延岡市、西都市、木城町、川南町、都農町、門川町、日南市*、串間市、国富町、綾町、美郷町、高千穂町、都城市、小林市、えびの市、三股町、高原町",
+            "福岡県：久留米市",
+            "佐賀県：神埼市、白石町",
+            "熊本県：阿蘇市、産山村、高森町、南阿蘇村、熊本市南区、熊本市北区、八代市、菊池市、宇土市、宇城市、合志市、美里町、西原村、氷川町、人吉市、多良木町、あさぎり町、芦北町",
+            "大分県：大分市、臼杵市、佐伯市、竹田市",
+            "鹿児島県：鹿児島市、霧島市、いちき串木野市、南さつま市、伊佐市、姶良市、鹿屋市、垂水市、曽於市、大崎町、東串良町、肝付町"
+          ]
+        }
+      ]
+    },
+    "images": [
+      {
+        "versions": [
+          {
+            "url": "https://www.asahicom.jp/imgopt/img/4ff96428f2/comm_L/AS20250113003419.jpg",
+            "query_width": null,
+            "size": null,
+            "type": "image/jpeg"
+          }
+        ],
+        "is_cover": true,
+        "description": "写真・図版",
+        "caption": null,
+        "authors": [],
+        "position": 737
+      }
+    ],
+    "publishing_date": "2025-01-13 21:37:00+09:00",
+    "title": "宮崎県で震度5弱、高知と宮崎に1メートルの津波注意報　気象庁",
+    "topics": [
+      "社会",
+      "災害・気象",
+      "宮崎県"
+    ]
+  }
+}

tests/resources/parser/test_data/jp/meta.info

@@ -1,4 +1,8 @@
 {
+  "AsahiShimbun_2025_01_13.html.gz": {
+    "url": "https://www.asahi.com/articles/AST1F4445T1FUTIL02SM.html",
+    "crawl_date": "2025-01-13 14:12:17.527262"
+  },
   "TheJapanNews_2024_10_13.html.gz": {
     "url": "https://japannews.yomiuri.co.jp/politics/politics-government/20241013-216478/",
     "crawl_date": "2024-10-13 16:27:01.520980"