are the default config for extracting text/tables the best ones?

[sergenti]

if I had to use only one table_settings, what should I use?
I'm working on a SaaS project and can't manually change values depending on the context

I'm dealing with the following types of documents:

• scientific papers
• legal documents
• standard financial documents (like 10k, 10q, s1, etc)
• other confidential financial documents (PE internal reports, transcripts, memorandums)
• etc

Some ideas?

right now for these types of document seems that the base pdfminer.six works best for text extraction, and that another library called tabula-py works best for extracting tables.

maybe I am missing something, this library seems so well-written.

[jsvine]

Hi @fylls, thanks for your interest and for the kind words. PDFs come in such a range of layouts, it's hard to say what the best settings would be. The default settings are supposed to be useful for tables that have clear, visible cell borders. But not all tables do have that.

As for text extraction, the same principle applies, but I think you have more flexibility in dynamically determining the best settings by (for example) examining character size, spacing, et cetera.

Open to other thoughts / suggestions on this, though!