search

poloclub / unitable

UniTable: Towards a Unified Table Foundation Model
https://arxiv.org/abs/2403.04822
MIT License
381 stars 28 forks source link

Is this possible with other language? #5

Closed yellowjs0304 closed 6 months ago

yellowjs0304 commented 6 months ago

@ShengYun-Peng Hi, Thank you for sharing your amazing work.

while running your pipeline, I have some questions. Does your model work with other languages? like Japanese, Chinese, Korean etc..

I think the "content.pt" model is the role of OCR. Is this right?

If I use the other OCR or PDF extractor like PDFParser, Is it can be applied to your pipeline?

ShengYun-Peng commented 6 months ago

Hi @yellowjs0304, thanks for your interest!

Yes, "content.pt" is used for OCR, i.e., table cell content recognition. I believe any OCR model that works for multi-line text recognition can be employed in UniTable's cell content branch. The beauty of UniTable lies in unifying all three tasks into one self-supervised training and inference pipeline, rather than combining existing methods from non-table domains.

yellowjs0304 commented 6 months ago

Got it, Thank you.