One mention two type
Closed this issue · 1 comments
iofu728 commented
After re-read the paper, I don't know how to evaluation multi-type? every type must predict?
In my mind, the hypergraph model can't do that thing.
In the error analysis, I find a strange result.
like
6,7 DATE|3,3 CARDINAL|7,7 DAY|1,1 INDEX|4,4 UNIT|1,1 NAME|3,4 PERCENT|1,1 MEDIA|6,7 EVENT
There are many mention links to two entity types.
I'm confused, It's not nested phenomenon.
It's 1612 mentions have multi-type in the test set.
Again, wish you guys best luck in your ACL submission. ; )
iofu728 commented
I understand the problem. There are also happen in GENIA.