The obvious solution is to actually have librarians correctly classify the videos. DDS focuses on the nature of the work itself, not on the keywords or spam in the content. Librarians understand how to class all kinds of works, and it should be relatively simple to build a DDS/MDS index (Melville Decimal System since it's open, see https://librarything.com/mds) for YouTube videos. Just like with books, disagreement on classification is inevitable and perfectly natural; there's no perfect classification scheme, though DDS/MDS does a generally good job.