Named Entity Recognition for GUM

Table of Contents

• Named Entity Recognition using Neural Neworks and Transformers Approach
• 1. Problem Statement
• 2. Data Description
  • Attribute Information
    • Inputs
    • Output
• 3. Topic Modelling
• 4. EDA
• 5. Modelling Evaluation
• 6. Results

Name Entity Recognition using Neural Networks and Transformers Approach

This dataset contains release versions of the Georgetown University Multilayer Corpus (GUM), a corpus of English texts from twelve written and spoken text types.The corpus is created as part of the course LING-367 (Computational Corpus Linguistics) at Georgetown University. Thus, our aim is to use two different kind of classifiers in order to accomplish the NER task.

1. Problem Statement

Classify correctly the 23 classes

2. Data Description

Data is obtained from this repo.

• Number of instances - 44111 entries (Train), 18236 entries (Test)
• Number of classes - 2

  Attribute Information

  Inputs

  • token: string feature

  Output

  • ner_tag : a classification label , 23 classes

3. Topic Modelling

The Topics are analyzed via two methods:

• Latent Dirichlet Allocation (LDA)
• Negative Matrix Factorization (NMF)

4. EDA

5. Modelling Evaluation

• Algorithms used
  • BI-LSTM
  • BERT
• Metrics used: Accuracy, Precision,Recall, F1-Score

6. Results