Custom Loss :

[3rdCore]

No description provided.

[3rdCore]

Hello,
I am currently using your package as part of an academic project, and I would like to know if you could help me to figure out how to implement a custom loss that will penalize mis-classification of certain tags.

I implemented a NER model for BILOU tagging (Begin, Inside, Last, Outside, Unit), but my dataset is quite unbalanced as it contains 80% of Outside tags, hence I wanted to penalize a mis-classsification of rare tags using that custom loss in order to force the model to pay more attention to rare tags (BILU).

Do you think that could be possible, as the crf_log_likelihood seems to be hardcoded and specific to CRF models.

Thanks for your time,
Tom