Comment by ryanisnan

Interesting. Do you have a uniform data set? E.g. documents of a specific type that you know consistently have similar formats, or is this training something you need to do per-document?