Multilingual Question Answering

Componenets

We have implemented two componenet in this:

1. Translation
2. Question and Answering

Objective

Helping people from regional areas, who are struggling to understand English. Through this project we are trying to answer questions which are asked in regional languages.

Models

1. LSTM attention model for Translation.
2. BERT Transformer for Question & Answering.

Metrics:

1. BLEU Score for evaluate translation.
2. Precision, Recall and Accuracy for Question and Answering.

Model Architecture:

Translation:

QnA Models:

Results:

Combined Model Results: