Ingest table (표) for better understanding table contents of LLM about ragchain

[vkehfdl1]

How about this?
https://mathpix.com/pdf-reader

[vkehfdl1]

Mathpix is too~ expensive OTL
https://nanonets.com/blog/table-extraction-deep-learning/
We can use table OCR model.

[vkehfdl1]

https://github.com/chezou/tabula-py
https://pypi.org/project/pix2tex/
Wonderful libraries

[vkehfdl1]

Using unstructured pdf loader might be okay

[vkehfdl1]

Maybe DIT + pdfminer.six method proposed in this paper will be great.

• change pdf to image using pdf2image
• create bounding box using DIT (from hf transformers)
• extract passages at bounding box using pdfminder.six

[vkehfdl1]

I implemeted nougat loader at #217 and @bwook00 is making deepdoctection loader.

[vkehfdl1]

We implemented Deepdoctection loader for recognizing table.
Using TableFormer or rest of thiings can be further implementation.
Adios