How to export parsing results from Ragflow in Markdown format

[Neo966]

I want to export the parsing results from Ragflow as a Markdown file and then compare the Markdown files to evaluate the parsing performance of Docling and MinerU. Can Ragflow do this?

[liuzhuosin]

I think what he might mean is simply converting the file to md and then comparing the results. The official documentation mentions that you can set MinerU's output directory, so you can check it out here: https://github.com/infiniflow/ragflow/blob/main/docs/faq.mdx#how-to-use-mineru-to-parse-pdf-documents

[Neo966]

@Magicbook1108 The table of pipeline parsing results is messy, but the naive one is fine?

pipline parse result：

naive parse result

[Magicbook1108]

The Naive Parser and the Parser in the Pipeline are distinct components in RagFlow.
You can review their code implementations in:

Naive Parser:
rag/app/naive.py

Parser in Pipeline:
rag/flow/parser/parser.py

[Neo966]

@Magicbook1108 Have you compared the parsing performance of DeepDoc, Docling, and Mineru? It seems that Mineru performs the best. What optimization configurations does DeepDoc offer?

[Magicbook1108]

DeepDoc uses OCR (Optical Character Recognition) to extract content from PDFs. If you provide an vision llm, then it will use vllm to extract information from figures.

[Neo966]

Do you have any recommended VLMs on your side?