Hi, I would like to ask if prodigy TextCat has the same feature like ner.correct, where I get text pre-labelled for me and I just correct it? textcat.teach does not seem to work fine for multiclass annotation and I feel like something like ner.correct would do better.
Hi! I think what you're looking for is the
textcat.correct recipe, which was introduced in Prodigy v1.11: https://prodi.gy/docs/recipes#textcat-correct
It lets you load in an existing trained text classifier and will pre-select the categories predicted by the model. You can then correct the predictions manually. The
--threshold option lets you configure the score required to consider a category as applying.
Yes, exactly what I am looking for, thank you so much!