From: Automating curation using a natural language processing pipeline
File type | Normalization | Total interactions | Correct interactions | % of gold | Estimated recall |
---|---|---|---|---|---|
Exact | 2,052,605 | 744 | 36.5 | 61.7 | |
HTML | Exact | 1,916,527 | 737 | 36.1 | 61.6 |
Fuzzy | 17,583,994 | 1,002 | 49.1 | 66.7 |