It works best with any plain ASCII text. What I've been doing is taking PDF's and using `pdftotext` and `iconv` to convert documents into plain text. I'll add more information on this in the README.md to make it more clear what it reads.