Train on messages received in Facebook chat and remember your emotional state depends on message. Next replies with chatbot based on it.

To download your Facebook data you need to click you profile and choose Download a copy of your Facebook data below your General Account Settings. Next wait for a link in email.

To run notebook you'll need Python3, Tensorflow and

pip3 install -r requirements.txt

You need to setup paths to download data and usernames which used in conversation.

If you want a chatbot (latest cell), you also need setup FB page, API keys, SSL and async call model.

The model is based on Convolutional Neural Networks for Sentence Classification by Yoon Kim with static word2vec embedding trained on text received. It performs over 90% on IMDB 50000 movie reviews, however for recall your emotion statement it needs to be trained on big number of conversations.

There are some problems with other existing Keras implementations - some of them use deprecated Graph API, some don't use external static word2vec, some limited to Convolution1D, single category, most did not ready for call via web service.

Possible improvements : add learning from aditional parameters like text replied, external word2vec or glove, mixed emotion state, Redis for async call.