There is a system which contain conversation text which is between two or more people, I just need to extract question’s answer from it, if it is discussed in conversation, through python nlp process. let say conversation text is as follows:
 

A: What are you going to eat with your sandwich?

B: I think I'm going to eat a piece of fruit.

A: What kind do you like?

B: I really like apples and grapes.

A: What kind of apples do you like to eat?

B: I love green apples.

A: I don't like green apples.

B: Really? Why not?

A: Green apples are too sour for me.

B: So, you like red apples better?

A: Yes, I love red apples.

B: I think green apples are a lot better.

And I asked question from that:

Which fruits discussed?

So it reply as

apple and grapes

or in other scenario it will show all conversation in which fruit is discussed.

I have checked many libraries but not able to figure out which library is appropriate or give desired result? I have to know, what step should be followed to achieve this functionality in python.

I have checked spaCy, and I found that it has vector and closest word functionality, but not able to figure out, how I can accomplish it based on the question. or any other library or packages which will help me on that.