search

infiniflow / ragflow

RAGFlow is an open-source RAG (Retrieval-Augmented Generation) engine based on deep document understanding.
https://ragflow.io
Apache License 2.0
10.08k stars 974 forks source link

[Question]: How to increase splits in the answer context #1203

Open lzivan opened 1 week ago

lzivan commented 1 week ago

Describe your problem

Dear Ragflow Team,

I am encountering an issue where the answers I need are split between the end of page 1 and the beginning of page 2. Currently, the Ragflow model only retrieves and uses the information from page 1, ignoring the relevant content on page 2. This leads to incomplete answers.

Could you please advise on how to configure Ragflow to handle such splits in the retrieved answers effectively?

Thank you for your assistance.

KevinHuSh commented 1 week ago

My suggestion is that:

  1. try to use other parsing methods like manual, laws, book.
  2. Increase page size per task( press the button on the right side of a specific document)
  3. Use RAPTOR.
lzivan commented 1 week ago

Thank you for your response. I tried the above methods, but they didn't solve the problem.

Additionally, I want to ensure that is it possible to make content split across pages get combined into a single split？

KevinHuSh commented 1 day ago

Thank you for your response. I tried the above methods, but they didn't solve the problem.

Additionally, I want to ensure that is it possible to make content split across pages get combined into a single split？

Change the page size per task to 1, and use 'General' and set the chunk number to a large number.