Links to the implementations of neural conversational models for different frameworks. Contributions are welcomed.
Keras
The most popular implementation of Seq2seq archintecture on GitHub. However there is still no evidence of getting good results with this repo.
A wrapper for farizrahman4u/seq2seq used for running experiments. No good results were achieved so far.
Need to check this up.
Lasagne
The model manages to learn language on syntactical level:
what do you think about bill gates ? → you ' re not a little man , baby . .
what happens if machines can think ? → i ' m going to be a lot of your wife
what is the greatest novel every written ? → what are you doing ?
have you hurt anyone ? → i can ' t believe it .
Torch
Probably the best results currently achieved with an open-sourced Seq2seq implementation:
Hello? → Hi.
How are you? → I'm fine.
What's your name? → It's hard to describe.
How so? → I'm not sure.
What color is the sky? → It's blue.
What is your job? → It's not that i'm a fucking werewolf!
Tensorflow
Modified code of machine translation model. No answering randomisation is implemented in this code, so the models answers with the same phrase way each time:
hello baby → hello
how old are you ? → twenty .
i am lonely → i am not
nice → you ' re not going to be okay .
so rude → i ' m sorry .
are you a robot or human? → no .
are you better than siri? → yes .
No good results so far.
Give it a try.
Get a lot of raw movie subtitles (~1.2Gb)