Multilabel Confusion Matrix non-binnary labels

,
1

Hello Prodigy community.
I have started using Prodigy quite recently.
I have annotated and trained my model. To get a broader perspective on the model's performance I am generating the confusion matrices for each label individually using the sklearn library. It would be interesting to create the multilabel confusion matrix from the whole model. Nevertheless, I cannot manage to get it done.
Is there any custom functions that might help me?
The results I want to end up with look something like this:

image

Thank you for your help

2

Hey @Alberto - thanks for reaching out!

Unfortunately, we don't currently have helpers to generate the multi-label confusion matrix for a whole model as you described. However, there's been some discussion on how to obtain a confusion matrix from spacy evaluate - there might be some answers there that could help you out.

We're currently working on a prodigy evaluate functionality so do watch this space!