I am unsure how to solve this problem. I have a set of terms I created using a custom model. I am now training a model with my terms, but sometimes my terminology from the list isn't being highlighted. This is during the same training sessions. So, in one instance, "meconium" is highlighted, but in another instance, it isn't:
Any ideas what I might check to solve this problem?
Hi! Which recipe are you using for your workflow? If you're annotating with a workflow like
textcat.teach, keep in mind that it will mix suggestions from your patterns with predictions from the model. The patterns are mostly there to help pre-select enough candidates in the beginning so the model starts off seeing enough positive examples that are relevant for your label. Later on, you will keep seeing more suggestions from the model you're updating in the loop.
So it's possible that the examples you're seeing where the terms aren't highlighted aren't based on pattern matches and were instead selected by the model. If you look in the bottom right corner of the annotation card, you should either see are
score, or a
pattern – this indicates whether the suggestion came from the model with a given score, or from one or more patterns.
Ahhh! Thanks so much -- that is exactly what is going on.