interpreting ner.train results

, , ,
1

Hi all,

I am training my model using

!python -m prodigy train --ner tweetsdb

and getting the attached results:

grafik
grafik742×797 28.7 KB

My question is:
How do i intepret the results given that I have 2 NER labels. My classess are not balanced and I am worried that my model could be correctly identifying the entity with the highest number of examples but not the other...

2

Hi! This is definitely reasonable and you can easily check this by setting the --label-stats flag when you train your model. This will show you a breakdown of the scores per label, so you can see how the individual labels perform.

3

Worked like a charm.

Thank you

1 Like