Cornell Movie-Dialogs and similar corpora derive from copyrighted scripts and state no dataset license, while the CMU corpus is CC BY-SA from Wikipedia text. For an open hub, which corpora can be hosted, which only linked, and what is the position on training use?
Why this is still open: Pages read give no dataset license for Cornell or WritingPrompts; the CMU license has no version number; legal position on training use varies by jurisdiction and was not assessed.