This project demonstrates how to train and use a Named Entity Recognition (NER) model using PyTorch and the Hugging Face Transformers library.

• Python 3.11
• transformers==4.28.1
• PyTorch
• restaurant-model/ directory (pre-trained model)
• place pytorch_model.bin (https://drive.google.com/file/d/1UMXelrj96fUBpnnhaQgRDdmXQ3-Yc-KH/view?usp=sharing) in restaurant-model/

To use the pre-trained model, run the following command:

python test.py -i 'how many 5 star restaurants are there in my area'

The output will display the entities found in the input text, along with their entity group:

Entity Group : Rating, Word :  5 star
Entity Group : Location, Word :  in my area

To train a NER model using your own dataset, you can modify the train.sh script. Here's an overview of the steps involved:

1. Prepare the dataset
2. Tokenize the dataset
3. Define the model
4. Train the model
5. Evaluate the model

The resulting model can then be saved and used for inference.

This project uses the restaurant dataset from the MIT Conversational Modeling Toolkit, which is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.