I saw this patent_inventor dataset given by PatentView, unfortunately no official document of how they generate this is found. I examine it by trying to merge back with disambiguous version datasets patent, inventor, location respectively. I assume since the later are where the information comes from, when merge with the latters as using dataset, the unmatched observation should all comes from the latter datasets. But when I merge patent_inventor(master) with patent(using), both dataset contain unique observation unmatched. This also happens when I merge patent_inventor(master) with location(using). I wonder why this would happen? Is there any mistake?