Communication between chunks/paragraphs

[corlay-MERCE]

In this version of Augmentoolkit, each chunk/paragraph is treated independently, correct?
The questions/answers generated from one paragraph cannot benefit from the knowledge contained in another paragraph (or from answers generated from other paragraphs). One can for instance think about a scientific paper where important definitions are provided only in the first paragraphs.

What about using the questions/answers generated for paragraph n-1 (as a summary of paragraph n-1) along with paragraph n? (But then the paragraphs could no longer be treated in parallel...)

[e-p-armstrong]

Correct, paragraphs are independent. This is partly for simplicity and partly for async processing. However you could make a pipeline that handles things as you describe.