How does dockling compare to unstructured.io?

[athrael-soju]

I've used unstructured.io for quite some time now and their open source docker API is decent. A few months ago I did some analysis on it: https://athrael.net/blog/experiments/brief-analysis-on-rag-with-pinecone-and-unstructured-io

I was wondering if anyone's ran any benchmarks on how dockling compares? I'd love to do a similar analysis for it, but I've just not had the time recently.

[dolfim-ibm]

Comparison of output and performance is coming soon with various systems.

[dolfim-ibm]

We have comparison results in the latest publications which just got accepted yesterday for a workshop in AAAI. We have some issue on arxiv.org, but we will post the paper temporary somewhere else.

[simjak]

Could you post here as well 🙏

[dolfim-ibm]

While we clear up the situation, here is a copy of the paper #574.

[KentonParton]

Thanks for sharing the paper @dolfim-ibm and congratulations on the release!

I enjoyed reading through it and found the speed comparisons with other libraries very informative. I'm curious - do you have plans to include quality/accuracy comparisons in future evaluations? It would be interesting to see how docling's output quality compares.

[cau-git]

@KentonParton Yes, quality comparison is a work-in-progress at this moment. We will share updates on this in a future iteration.