-
"datasets" folder contains superclean_controlled.csv and superclean_depressed.csv, which are already cleaned, for twitter and reddit_depression_suicide.csv which are from reddit.
-
"MLM" folder is for KE_MLM model processing and to save the model.
-
LIWC tokens processing are saved in "tokenizers".
-
"reddit_baseline", "reddit_liwc" and "reddit_mlm_ke" are for the baseline Distilbert_base_uncased model , model with added LIWC tokens and knowledge-enhanced model with masking respectively.
-
"twitter_baseline", "twitter_liwc" and "twittwe_mlm_ke" are the same with reddit's part.
-
"runs" is for the saved log and "weights" is our trained weights.
-
"BertDataset.py" is for customDataset class and "logger.py" for tensorboard things.
-
"PreprocessingCombined.ipynb" is for the data preprocessing(http removal, non-english word removal, etc)