Parser comparison

[alensiljak]

It would be interesting to have some parser comparisons for documents.

I have been comparing Upstage to Gemini, run directly in AI Studio. Upstage was fairly poor in parsing the lab results of this particular document.
Gemini did a better job overall but also quite imperfect. Many values were simply missing. However, when pasting an image of individual page from the document, the results were excellent.

I will check out how to use Docling directly and compare the results.

[alensiljak]

Just saw
https://github.com/Goldziher/kreuzberg

[pu-re]

how was it?

[alensiljak]

So far, Gemini was great in parsing individual pages so I only used that. Did not have time yet to try any other option out there.