Improve cleaning based on semantic type
Closed this issue · 3 comments
uk.gov.dstl.baleen.annotators.cleaners.helpers.AbstractNestedEntities will merge based on the first entity found (or least confidence).
Perhaps it should also consider the semantic type, a more specific type (eg Entity vs Person) should pick the person (for the same confidence)
I'm not sure if it will merge two entities of different types, I think they have to be the same?
Sorry that's correct. I was thinking this would be an additional mode / special case for two entities which are subtypes of one another.
Perhaps the current type system isn't complex enough to warrant it though.
Possibly something to think about in the future. As you say, I don't think our type system is complex enough at the moment for this to be an issue.