-
Notifications
You must be signed in to change notification settings - Fork 13
Open
Labels
Description
@kleinay
Shany mentioned that we get lower scores on entity recognition - probably since we group noun compounds.
I think we should:
- Get her evaluation code somewhere on github (probably first to her fork and then PR to Vered's). BTW, I don't think she's a member of the OKR project.
- Run the evaluations to get the lower numbers on entity identification
- Decide what kind of relation we want between head of noun compounds and their dependent words, see this dependency parse, for example
- Re-run the evaluations and see if the numbers improve.