A repo to try to generalise abusive language detection
All data resides in the data folder. And all code in the src folder.
Further, the source folder is subdivided into different folders, where the top-level folder only contains an init.py file and a single file to run all models (which is run with options selecting dataset, model, parameters, etc.) while each individual model is contained within its own folder, and finally there is a folder containing shared resources such as feature generation scripts, data reading, etc.
Represent sentences as:
- LIWC
- Word length
- Word complexity
- Sentiment
- POS (PTB)
- Use Davidson, Waseem, Waseem & Hovy to show adaptability.