Vector embeddings are approximation engines that are excellent at finding semantically similar content, but systematically weak at distinguishing specific entities ...