semantic-f1-score

Semantic F1 semantic-f1-score is a drop-in replacement for sklearn’s conventional f1_score in subjective or fuzzy multi-label classification. It keeps the familiar precision-recall framing while using a domain similarity matrix to acknowledge when wrong labels are still semantically close. The package is the reference implementation accompanying the paper: ‘Semantic F1 Scores: Fair Evaluation Under Fuzzy Class Boundaries’.

To install: pip install semantic-f1-score