The PaperMage documentation claims that all Entity will have both a list of Spans, i.e. a list of indices into the original document text, a list of "boxes" which point out where that entity exists on a page. This issue consists of a number of different things:
[ ] verify that we can extract bounding boxes from entities like "tokens" and "words"
[ ] Once we have entities for annotations from #2, intersect them with tokens in the original text
[ ] Once we have entities from information extraction, cross-reference those to the highlighted annotations.
The PaperMage documentation claims that all
Entity
will have both a list ofSpan
s, i.e. a list of indices into the original document text, a list of "boxes" which point out where that entity exists on a page. This issue consists of a number of different things: