The models are trained on data from the Switchboard Dialogue Act Corpus. http://compprag.christopherpotts.net/swda.html
Must download NLTK in order to run the pos tag parser.
They are named according to the feature set within them. For example, the LIWC jupyter notebook file contains a classifier that uses only LIWC features. The combined_feats+LIWC has all three unique feature set models with LIWC features concatenated.
The notebooks classify the data set if run in the order presented.
The csv files all hold the features for their respective models. They are named after the jupyter notebooks that created them.
The helper functions in the files are documented according to their purpose. Comments throughout the jupyter notebooks clarify the programs making them easily understood.