Skip to content

A NLP RNN trained on CR transcripts and optimized for Azure

Notifications You must be signed in to change notification settings

Berylbell/CR_NPL_RNN

Repository files navigation

Contained within this repository are the tools required to train a recurrent neural network from the TensorFlow library on transcripts pulled from the Critical Role wikia. The scripts to complete this project from scratch are included. The pullTranscripts.py file will pull the transcripts from the appropriate source, and the trnscriptClean.py file will clean and preprocess them for use in the neural network.

After the transcripts have been downloaded the rnn can be run on them. The network is structured to be run on the text of each speaker individually to learn their speech patterns. A rnn is also trained on the most common speaker order. The text generated is in the learned order, with a topic being passed from the previous speaker to the next, in order to simulate the flow of the conversation.

About

A NLP RNN trained on CR transcripts and optimized for Azure

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages