textcat.eval without manual labeling?

Hi - I have a question about an evaluation set. I’d like to look at individual false negative and false positive examples to get a sense of what’s going on. Is there any way I can get the score in the output? Or, is the idea that I would just pass the examples through SpaCy directly to get the score?

I think you might want to just write a custom recipe that passes the data through spaCy and then displays it in a format that’s convenient to you. We’ve tried to strike a balance in the built-in recipes between covering basic workflows, and giving the user less to learn.

Personally I find when I’m using programming tools with a large interface surface that I often would prefer a smaller number of composable primitives. That’s why we’ve focussed on the Python API a lot, rather than trying to do everything through the command-line API. Ultimately if you’re composing command-line commands, you end up writing bash scripts, which ends up being a worse developer experience than doing the same thing in Python.

You can find a good set of custom recipes to use as a starting point in the custom recipes repo: https://github.com/explosion/prodigy-recipes .