chatbot-rnn

A toy chatbot powered by deep learning and trained on data from Reddit. Built on TensorFlow v1.4.0 and Python v3.5.1.

Here is a sample chat transcript (not cherry-picked). Lines that start with ">" are me; the lines without ">" are the chatbot's responses.

$ python3 chatbot.py
Creating model...
Restoring weights...

> Hi, how are you?
 I'm not the only one here that's having fun with you, it's all me.
> Who else is with you?
 Haha. I'm not the one who made the comment. I'm the one who said that and didn't think that much of anything. 
> Oh, I see. So what's new?
 Nothing. I just thought it was a bit of an overreaction. It's been a while since I've read it. 
> It's all good
 It's all good. It's one of those things that makes me smile :)
> That's great. This is actually my demo chat, so it's nice that you're keeping things positive!
 Thank you :) 
> You're welcome! It's a nice change from the way that you sometimes are...
 *blushes* Sorry, I should have clarified that I wasn't clear. I apologize for my ignorance.
> Sooo, where are you from?             
 I am from South America. Where you from?
> California. Do you like it in South America?
 I live in Australia, but south of Florida is one of the reasons I live in Florida. I've never been there, but I've never been there. 
> I see, a true citizen of the world
 Nope, I've never heard of it. I just spent a lot of time there. 
> Spend a lot of time where?
 I've never been there before, but I've heard it's about an hour and a half away. Maybe there's some sort of party called "Southern" but that's about it. 
> Maybe there is, that's a great observation.
 Not to mention it's all about being from another country.
> All right, guess I should wrap this up, nice talking with you
 I'm not going to argue with you. You're doing it wrong.
> Some might call that arguing. Any final words?
 Don't do that. I don't need to. 

Get Started

Run my pre-trained model

Try playing around with the arguments to chatbot.py to obtain better samples:

These values can also be manipulated during a chat, and the model state can be reset, without restarting the chatbot:

$ python3 chatbot.py
Creating model...
Restoring weights...

> --temperature 1.3
[Temperature set to 1.3]

> --relevance 0.3
[Relevance set to 0.3]

> --relevance -1
[Relevance disabled]

> --topn 2
[Top-n filtering set to 2]

> --topn -1
[Top-n filtering disabled]

> --beam_width 5
[Beam width set to 5]

> --reset
[Model state reset]

Get training data

If you'd like to train your own model, you'll need training data. There are a few options here.

Once you have training data in hand (and located in a subdirectory of the data directory):

Train your own model

Thanks

Thanks to Andrej Karpathy for his char-rnn repo, and to Sherjil Ozair for his TensorFlow port of char-rnn, which this repo is based on.