Investigating Annotator Bias with a Graph-Based Approach

Install requirements.txt
Place the files attack_annotated_comments.tsv, attack_annotations.tsv, and attack_worker_demographics.tsv from the Wikipedia Detox Project in folder 00_data
Run 20_graph-generation.ipynb to generate the graphs with teh different weight functions
Run 21_cluster-and-split.ipynb to generate the data sets (train/test split) for the different groups and weight functions
Run 24_distillbert_classifier.ipynb to train the different classifiers

Outputs can be found in 03_results

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
00_data		00_data
01_processed_data/200903_run_004/datasplits		01_processed_data/200903_run_004/datasplits
03_results/200903_run_004		03_results/200903_run_004
helper		helper
20_graph-generation.ipynb		20_graph-generation.ipynb
20_graph-statistics.ipynb		20_graph-statistics.ipynb
21_cluster-and-split.ipynb		21_cluster-and-split.ipynb
21_visualize_graph.ipynb		21_visualize_graph.ipynb
24_distillbert_classifier.ipynb		24_distillbert_classifier.ipynb
README.md		README.md
requirements.txt		requirements.txt

Provide feedback