Document embeddings and the search for precise facts
A user argues that adding tourist information to a sentence about Paris can make its document embedding lose the match for a simple question: “What river is Paris on?”
TLDR
A user contrasts “Paris is a city on the Seine” with a longer passage that adds tourist attractions. They argue that the short sentence directly matches a question about the river. The longer passage would probably match Paris tourism queries, they say, but its embedding—the text’s representation for search—becomes muddled and loses the match for that narrower question.
Document embeddings and the search for precise facts
A user argues that adding tourist information to a sentence about Paris can make its document embedding lose the match for a simple question: “What river is Paris on?”