How to perform QnA based on Product Fact Sheet pdfs?

Question

I have lots of product factsheet PDF documents containing product specs in tabular format. Almost, all the documents are in a similar layout. Parsing the documents, chunking them and storing in to a Vector Database and querying them didn't help much. It was not able to capture the context. For ex, when I tried to search for the product category of a product, it returned irrelevant chunks. What is the best way to handle this use case?

Answer 1

Read the below blog and see whether it helps you getting started?

https://techcommunity.microsoft.com/t5/startups-at-microsoft/build-a-chatbot-to-query-your-documentation-using-langchain-and/ba-p/3833134