Text Extraction AKA Question Answering Using BERT

Performing Text Extraction also known as Question-Answering using BERT,and serving it Via REST API.

More about BERT can be found here
End to End NLP Text Extraction Probelm.
The Kaggle dataset can be found Here Click Here
My kaggle Notebook can be found here

Steps to Run the Project:

create a virtual environment and install requirements.txt

For Training

After Setting up the environment go to backend/services/text_extraction/application/ai/training/ and run main.py and the training will start.
After training is complete the weights of the model will be saved in weights directory, and this weights can be used for inference.

For Prediction/Inference

Download the pre-trained weights from here and place it inside the weights folder(backend/services/text_extraction/application/ai/weights/trained_weights)
After setting up the environment: go to backend/services/text_extraction/api and run app.py.
After running the above step the server will start.
You can send the POST request at this URL - localhost:8080/text_extraction/api/v1/predict (you can find the declaration of endpoint under backend/services/text_extraction/api/init.py )
You can also see the logs under (backend/services/text_extraction/logs) directory.

Following are the screenshots for the sample request and sample response.

Request sample

Response Sample