Hey ! Well, you can use this model for the Q&A model- given a context and question corresponding from the context, this model can answer the question.

Beware that this model can only answer closed-end questions. If you have given context based on the history of the USA from a Wikipedia page and asked the question 'Why the sky is Blue ?', chances will be that you wouldn't get the answer you expected.

The exceptional thing is I fine-tuned this model with only 50 data points (out of 10000) due to RAM issues. Even though with such a small dataset, this is working exceptionally well with 98% accuracy!

You can input up to 30k words (That's equivalent to 100-page book content). But, it takes time to pre-process (expected!).

Ask the question which can be answered from the context only. This field can take upto 1000 words as input.

This section provides you the answer from the above question. Note: Sometimes you may not get the right answer!

This number lies between 0 and 1, 1 being best possible answer, 0 being least possible answer.

I used BERT model for this task. I've explored different methods and finally arrived at this model. This is pretty good one!

PS: I'm working on solving Q&A for physics problems. If you are a big physics buff, please reach out to me to make this possible! (mail:manideepmaruthi.234@gmail.com)