Ask HN: Any PDF Benchmarks?
I am testing several PDF parsing libraries, and unfortunately, most of them have issues. For example, many struggle with non-English languages, while others can’t reliably handle tables. Having a standardized PDF benchmark would be incredibly helpful — at the very least, it would allow us to identify obvious shortcomings without needing to install and test each library individually.
Is anyone building such a benchmark? For example, markitdown from MS can't recognize Cyrillic text, when I started researching into it, I found that they use pdfminer.six under the hood and there is an unresolved issue with supporting languages. Docling is OK with tables, but fails with cyrillic text; marker-pdf is OK with tables, but it also fails with cyrillic text; What other pdf parser libraries exist? I am looking for preferably on-premise solutions, but if I won't find a reliable and accurate solution, I might consider cloud based solutions as well.