Add NLP to extract structure and discovery metadata, including hypotheses

[nuest]

(original idea by @chriskray and @simonscheider - just cleaning up our internal ideas list!)

The extractor could use Natural Language Processing to find structual information and discovery metadata in the text part of a workspace, most importantly in the main document.

dkpro: https://github.com/DARIAH-DE/DARIAH-DKPro-Wrapper and https://code.google.com/archive/p/dkpro-tutorials/

Potential targetet content:

• hypotheses
• results
• p-Value
• ...