How to improve the recall and F1 score in entity extraction using CRFEntityExtractor

Hi Guys, I’m creating my first bot using rasa. for the extraction of entities I’m using, CRFEntityExtractor. which give me good precision, but poor recall and f1 score, that’s mean my NLU model is very good to predict correctly the entities but on the other side, he doesn’t recognize all values; NB: I did some regex and lookup tables to improve the recall but without success. could you guys tell me which parameter I have to tune to improve the recall?

1 Like