unstructured VS llama_parse

Be careful with unstructured:

https://github.com/Unstructured-IO/unstructured/blob/d11c70c...

from: https://github.com/open-webui/open-webui/issues/687

I’ve been looking at this

https://freeling-user-manual.readthedocs.io/en/v4.2/modules/...

at the freeling library in general, also spaCy and NLTK. The chunking algorithms being used in the likes of LangChain are remarkably bad surprisingly.

There is also

https://github.com/Unstructured-IO/unstructured

But I don’t like it, can’t explain why yet.

My intuition is that 1st step is clean sentences and paragraphs and titles/labels/headers. Then probably an LLM can handle outlining and table of contents generation using a stripped down list of objects in the text.

BRIO/BERT summarization could also have a role of some type.

Those are my ideas so far.

Unstructured is the best one I’ve used so far: https://www.unstructured.io

https://www.unstructured.io/ can parse PDFs, then you can feed all of them to Claude, which has a 100k context window.

If the copy & pase method does not work: One approach is to use the functionality of Unstructured to parse the PDF. If need be, it can do OCR on the PDF too if you have Detectron2 installed. After conversion you would still have to save it as an excel file though.

I would check out https://github.com/Unstructured-IO/unstructured (what lang chain uses) or https://github.com/axa-group/Parsr (probably what unstructured copied to get their startup off the ground lol)

It's fair to think that undesirable artifacts and lack of structural context would impact search accuracy, performance, and ultimately cost. Consequently, it makes sense to perform some data pre-processing before passing the source documents to the RAG workflow. Third-party APIs and tools, such as LlamaParse and LayoutPDFReader, can help with pre-processing PDF data, however keep in mind that source documents may take any forms and there is no one-size-fits-all solution. You may have to resort to developing custom processes for pre-processing and search your unique data.

I'm part of the team that build LlamaParse. It's net improvement compare to other PDF->Structured Text extractors (I build several in the past, includig https://github.com/axa-group/Parsr).

For character extraction, LlamaParse use a mixture of OCR / character extraction from the PDF (it's the only parser I'm aware of that address some of the buggy PDF font issues, check the 'text' mode to see raw document before reconstruction), use a mixture of heuristic and Machine learning models to reconstruct the document.

Once plug with a Recursive retrieval strategy, allow you to get Sota result on question answering over complexe text (see notebook: https://github.com/run-llama/llama_parse/blob/main/examples/...).

AMA