My target is to build a system that can accept a story and then able to give the answer of the question against the story.
I have used gensim and tfidf to achieve this goal. Also used some text similarity matrix as well for the task. I have got a decent accuracy on this when question are straight forward. However the main challenge is related to the sentence boundary detection on which I am still working. Next I will use core nlp module to convert sentences to the answer format(this part I am still working)
You will find it here:
https://llamasearch.wordpress.com/watch-demo/nlp-comprehend-de...
Please give your suggestion how can I improve on this and what other thing I need to do. If you have any better approach pleas share.
Do post here your experience and let me know about the model.
Note: Make sentence boundary well defined using .(full stop) then you will get best results.
Have fun :)
Comments