Implementing text categorization, CATS_SCORE always equals 100 and SCORE always equals 1

[fionaychen]

Hello!

I'm using spaCy to implement a text categorization task. When I run the command to train the model, I get a somewhat strange output. Specifically, it looks like the CATS_SCORE is always 100 and SCORE is always 1 (but the LOSS does change). I am quite new to using spaCy so I've been having trouble figuring out why this is the case, but it does not seem correct. Does anyone have a sense of why this might be the case?

E # LOSS TEXTCAT CATS_SCORE SCORE

---

0 0 0.25 100.00 1.00
0 200 43.38 100.00 1.00
0 400 44.46 100.00 1.00
0 600 40.68 100.00 1.00
0 800 31.62 100.00 1.00
0 1000 37.17 100.00 1.00
0 1200 34.96 100.00 1.00
0 1400 34.13 100.00 1.00
0 1600 32.05 100.00 1.00

To provide a bit more detail on the text categorization task, I have a dataset of the texts of many tasks, which I am trying to classify as either "promotable" or not. I am using the multilabel categorization with just 1 label.

I have tried to check a few things to fix this so far: it does seem like my dataset is set up alright (with ~20% of tasks being promotable) and there is not any overlap between the datasets I am using to train/develop/test the model.

Thank you for your help! And I am happy to provide more details about the code, etc as needed.