Seq2Seq-Chatbot

Seq2Seq Chatbot Using Tensorflow.

Dependencies

• Tensorflow
• Numpy
• Six

Usage

To preapre the enc and dec file for your own data use the pre_processing.py inside the data dir.

1. Start training the model by changing the mode = 'train' in bot.py.
2. Model will be saved in the checkpoint dir every few steps based on the value assigned to steps_per_checkpoint.
3. After training is finished set mode = 'test' and execute the script.

Triggers

If you are training a closed domain chatbot then you can use the trigger.py to make the chatbot perform an action based on the decoded output.

• Place the triggers in the dec file and make sure the trigger symbol is in vocab20000.dec.
• Update the trigger.json.
• Update the conditional statement and the action to perform in trigger.py.
• Train and run your model.

Other Files

• telegram.py - Just assign the API key to the bot var to interact with your bot on telegram.
• debug.py - Prints everything.

References

1. A Neural Conversational Model
2. Tensorflow Sequence-to-Sequence Models