Accurate text generation

I have a chat application that works with predefined messages. The database contains about 80 predefined chains, each of which contains 5 possible answers. To clarify, here is an example:

Q: "How heavy is a polar bear?"

R1: "Very heavy?"
R2: "Heavy enough to break the ice."
R3: "I don't know. Silly question."
R4: ...
R5: ...

Say the user selects R3: "I don't know. Stupid question."

Then this answer will have 5 possible answers, for example:

R1: "Why is that silly?"
R2: "You're silly!"
R3: "Ugh. I'm done talking to you now."
R4: ...
R5: ...

And each of these answers will have 5 possible answers; after which the conversation will end, and you will need to start a new one.

So, to remind you, I have 80 manual letters, each of which has 5 possible answers, going to 3 layers with a depth = 10,000 messages.

: , ?

RNN: RNN. RNN , .

, http://getvene.com/ .

+4
2

, , . Python Keras ( LSTM- TensorFlow). . , Gutenberg, . . , , , RNN, .

question responses. , . , , , "" . - , , - (), , - (). , - , - Document Distance ? , , - , / , .

+3

Source: https://habr.com/ru/post/1679155/


All Articles