Implement matching for CCN. This was 3 rounds of 4-people matches. Uses sentence transformers for embedding and cosine distance for comparing matches. See this post for background.
Run these in sequence:
python create_match_matrix.pypython get_semantic_scholar_info.pypython do_matching.pypython generate_word_clouds.py
You will need a data file, which contains PII, hence is not included in this repo; email me for a scrubbed version. Requires my fork of paper-reviewer-matcher. For GPT-3 based keyword inference, write the key (starting with sk) into .openai-key. For wordclouds, I recommend using the Fira Sans Condensed font, which you can download for free in TTF format.