I am fairly new to prodigy. But I have a great user experience. That said I have leveraged your tool for NER and Text classification. I would like to understand on how to interpret the output of ner train recipe ! I have attached a screenshot of one of the outputs here. NER_EN_2_t5|690x328
Thank you and look forward to hearing from you!
Glad to hear Prodigy is working well for you!
The NER training recipe outputs some simplified accuracy statistics. Specifically, accuracy refers to:
(number of correct predictions) / (number of total predictions + number of missed entities)
If you need to customize the reporting, you can either make a custom recipe with the extra accuracy logging you need, or you could also export the data and train with a different software package, for instance you could use
spacy train, which outputs slightly more detailed stats.