This endeavor could be automatic by ingesting sample metadata into an LLM and getting it extract enriched metadata. We assume this fun