Hi everyone, i have implemented ELECTRA model for a token classification usecase.
I am facing few issues in this.
- if ee is classified as ‘b-person’, then seeds is classified as sb-personds. ie model is identifying as s(ee)ds–>s(b-person)ds
- the model is identifying common nouns such as user and patient as proper noun and thereby classifying it as b-person.
Please advise any approach to handle these issues