README.md

Named Entity Recognition using a Bidirectional Long Short Term Memory (LSTM)

Model

• Dataset is provided by kaggle
• model build using Keras Deep learning API, implemented in this kaggle notebook

Usage

curl -X GET "http://127.0.0.1:8000/classify/Ali%20is%20swimming"

• response

{
  "Ali": "I-Person",
  "is": "O",
  "swimming": "O"
}

• The API is hosted on a heroku instance for easier access
• The API is implemented using FastAPI providing out-of-box documentation, checkout autogenerated docs in heroku

Local Development

• Install the requirements pip install -r requirements.txt
• start the development server with hot-reloading uvicorn main:app --reload

Directory structure

├── main.py             => main program file with API implementation
├── NER-BiLSTM.h5       => trained tensorflow model
├── Procfile            => heroku service file
├── README.md           => ...
├── requirements.txt    => required pip packages
└── word_idx.obj        => list of all words in used corpus