Maybe the solution is to have LLM read all that and summarize? And of course LLM can detect duplicated content - even if it is rewritten, it should generate a

Maybe the solution is to have LLM read all that and summarize? And of course LLM can detect duplicated content - even if it is rewritten, it should generate a similar meaning vector.