gpt4-pdf-chatbot-langchain VS unstructured

I am playing around with this github project, which takes a user question as input and immediately runs a vector search on it to find relevant storied information before delivering an answer.

I am using this https://github.com/mayooear/gpt4-pdf-chatbot-langchain as a reference to ingest PDFs into pinecone and chat with a document, but my results aren’t good. Since it’s looking for related documents, there’s no good relation to the meta question: “What questions were asked in this interview?”

Anyone can already do this locally with their own API keys for free, with no technical knowledge by cloning a github repo (e.g. https://github.com/mayooear/gpt4-pdf-chatbot-langchain - this one can also chat with multiple pdfs which is much better). Even with gpt-4, I just don't find the responses useful usually. I find the model doesn't do great with scientific stuff aside from asking very basic things. Might have to wait for gpt-5.

https://github.com/mayooear/gpt4-pdf-chatbot-langchain this repo uses gpt-3.5/4 which uses OpenAI API. Is there any work donw with free/open-source LLMs

Follow this guide https://github.com/mayooear/gpt4-pdf-chatbot-langchain

I am using this one on windows 10. Took 2 evenings to set up: https://github.com/mayooear/gpt4-pdf-chatbot-langchain

Thanks for your take on the subject. I agree that starting from scratch would be too much. I think my post above might be misleading in regard to training. I wouldn't suggest to start from scratch but to provide additional data to a pretrained AI. But you can use GPT-4 (through API) in combination with pinecone to provide data. Here is a project, where someone implemented this to work with large PDF files. I don't think it would be too hard, to start from there and adapt the project to the requirements of OP. Obviously this would require paid for API keys. LLama could be also a good starting point, with a lot of resources available.

I'm currently developing a chatbot application that interacts with PDF documents using GPT API, Langchain, and a Pinecone vector database. The project is built on this repository: mayooear/gpt4-pdf-chatbot-langchain.

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)