PRODUCTION SCENARIO
A company's support assistant must return policy passages that answer a user's question even when the passage and the question have few words in common.
Which approach will meet these requirements?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Index the passages as embeddings and retrieve by semantic similarity.
An embedding model turns each passage and each question into a vector, so the retriever compares them by meaning and returns a passage that shares no wording with the question. Keyword scoring ranks by term overlap, which is the failure described. A synonym dictionary has to be written by hand and still misses phrasing nobody listed. A fine-tuned model answers from memory and returns no passage to show the user.