Entities Mode: Sentence or Utterance Metadata for Co-reference Disambiguation

[Abhishek-P]

This is probably relevant only for multi-party dialogue like cases such as Podcast Genre.
In such cases, the same forms often switch back between different speakers and makes it hard to distinguish the specific parties

[amir-zeldes]

That's an interesting idea, it could be very useful for coref on spoken data. There is actually already a facility for showing sentence annotations on hover over the sentence number in sentence-by-sentence mode, I think you'll see it outputting the sentence type right now. The widget is aware of the annotations of the sentence, so that is not hard to display, but it's not aware of higher structures, and unfortunately the <sp> tags can contain multiple sentences, so right now those can't be used. If you know js and you're interested in doing some GUI work on GitDox I'm happy to point you in the right direction though 😉

[Abhishek-P]

Cool! Makes sense.
Let me try and take a pass at this.