An example project for named entity recognition(NER)

Approaches

1. Scratch Dense NN (Tensorflow + keras)
2. Long Short Term Memory (Pytorch + lightening)
3. Spacy
4. Scratch Transformer (Pytorch)
5. Transformer Fine Tuning (Hugging Face)

Result

Detail result analysis is available in each project directory.

We couldn't complete the following approaches due to time constraints

• 2.Long Short Term Memory (Pytorch + lightening)
• 4.Scratch Transformer (Pytorch)

For the other three approach the summery of the results are given below.

Dense Neural Network Model Overall score

Dataset	Accuracy	F1 Score
Train Set	0.8593	0.8579
Validation Set	0.8358	0.8685
Test Set	0.8392	0.9547

Spacy Transition-based Parser Model Overall score

Dataset	F1 Score	Precision	Recall
Validation Set	0.8554	0.8685	0.8426
Test Set	0.8658	0.8810	0.8510

DistilBERT Fine-tuning Model Overall Scores

Dataset	F1 Score	Precision	Recall	Accuracy
Train Set	0.5844	0.5833	0.5856	0.8844

Project File Structure

• Each approach have a directory associated with it
• These directories are python project themselves
• They have the following file structure
  • src directory (Contain all source code)
    • main.py (Starting point of the project)
  • explore.ipynb (Contain some exploratory code specific to that project)
  • requirements.txt (Contain the requirement of that project)
  • readme.md (Contain any specific command or note regarding that project)

Upcoming

The purpose of the server directory is to enable easy inference of all projects by serving them in a centralized location

Contribution

Thank you for considering contributing to our project! We appreciate your time and effort.

To start contributing, follow these steps:

1. Fork the repository and clone your fork.
2. Create an issue in the repository or choose an existing one.
3. Create a new branch for your changes (use a descriptive name with the issue name, such as #2-add-new-feature, #3-add-new-approach).
4. Make your changes in the new branch.
5. Test your changes thoroughly.
6. Commit and push your changes to your fork.
7. Use the issue number like #2, #3 ... in your commit message
8. Create a pull request.