grobid icon indicating copy to clipboard operation
grobid copied to clipboard

not doing well on recognizing references in footnotes

Open fredzannarbor opened this issue 2 years ago • 1 comments

Hi,

I have a 316 page PDF document about space warfare strategy with 342 footnotes, most of which contain references. I don't know the exact number, but most of those should be references -- 200 or more. Grobid is only finding 50-60 references. I see in #839 that finding references in footnotes is a known weak spot. That was in October 2021. What's happening now, and what are some strategies I could use to improve detection?

Fred

fredzannarbor avatar Jun 30 '22 05:06 fredzannarbor

Hi @fredzannarbor !

Nothing happened on the topic since last October. There's very few training data for references in footnotes at the moment and the normal approach would be to add training data to cover at least minimally this case.

kermitt2 avatar Jun 30 '22 18:06 kermitt2