An academic project exploring the use of convolutional neural networks and variants of word embeddings to detect abusive language in online text.

The current state of this repo is only the first step in the beginning to organize the material for open-sourcing. The thinking is to just get it committed, then organize and refine as the project wraps up.

• Add doc comments to top of files.
• Update GC-Commands-Reference.txt to properly represent commands used in the project.
• Flesh out a README for each of the directories in the project.
• Move the notebooks into the appropriate directories.
• Clean up the tfidf directory in order to include it in the repo.

All notable changes to this project will be documented here.

• GoogleNews-related files and directory to data directory.
• Jigsaw-related files and directory to data directory.
• PNGs of diagrams used in paper.
• Lexicon-related files and directory to the data directory.
• Specialized embeddings-related files and directories.
• Wikipedia Personal Attacks corpus-related files and directory.

• Moved old Gao feature extraction script to deprecated directory.
• Moved experiment scripts to experiments directory.
• moved runner scripts to runners directory in experiments directory.

• Docs directory.
• Diagrams in the docs directory.

• Renamed runner scripts to example_SCRIPT_NAME.sh.
• Abstracted runner scripts to be examples.
• Fixed variable references in runner scripts.
• Removed dependence on logger from cnn_text.py.
• Moved old versions of networks and experiments to deprecated directory.

• Experiment scripts.
• Deprecated scripts.
• Preprocessing scripts.
• Local runner shell script.
• Google Cloud Service runner script.
• Gitignore file.
• README file in each directory.
• Jupyter notebooks with related work.