Given that we might want to add in temporal information at a later stage, it could be helpful to have the entire text together at that stage,
just so we get any potential references/time anchors, which Heideltime could use to determine actual normalized results.
Best would be to check with a quick benchmark on sample data, to see how meaningful the difference would be.
Schematically, it is much easier to determine the actual paragraphs first, and then individually process them.
Given that we might want to add in temporal information at a later stage, it could be helpful to have the entire text together at that stage, just so we get any potential references/time anchors, which Heideltime could use to determine actual normalized results.
Best would be to check with a quick benchmark on sample data, to see how meaningful the difference would be. Schematically, it is much easier to determine the actual paragraphs first, and then individually process them.