Find answers from the community

Updated 2 months ago

What is the best set of instruments I

What is the best set of instruments I can use for the task of comparing everything vs everything in one PDF vs another PDF? And list down all the contradicting information (e.g.:
Plain Text
1) in PDF-A it is written that Microsoft made profit in 2021, but in PDF-B it is written that Microsoft made loss in 2021

2) Based on PDF-A, Apple spends more money on design than on technology, however in PDF-B it is stated that Apple spends more money on techonlogy etc.

)??

You see? I don't have any "initial query". I just want to compare the whole PDF-A vs the whole PDF-B and list down all the contradicting information.

I'm reading the LlamaIndex docs and frankly I have a headache now. Can't figure out which exact instruments to use and how.

use agents? ok, which?

use multi-hop query engine like in the example with Tesla and another company (forgot the name)? But I don't have initial query. Do I need to ask "compare everything and list down contradictions?"

Stuff as much as the model can handle, ask it to summarize then compare summary vs summary? Maybe? But summarization might omit important info...

etc.

Help, please
W
p
2 comments
Maybe using a LLM with larger context and add content from both the PDFs and compare straight forward
@WhiteFang_Jr Yeah, I am thinking that it is the best way in terms of accuracy... But I need to send a lot of texts... The pdfs are around 500 pages long
Add a reply
Sign up and join the conversation on Discord