Semantic search
Search that ranks by meaning rather than literal word overlap — useful when the corpus's vocabulary isn't your query's vocabulary.
Last updated
Search that ranks by meaning rather than by literal word overlap. The query is embedded into the same vector space as the corpus chunks, and matches are found by vector similarity. It finds passages about the same idea even when they share no words with the query.
Why it matters for your research. Semantic search is indispensable when the vocabulary of the corpus is not the vocabulary you came in with — period spellings, Latinate jargon, translated phrases, euphemisms. It is also less precise than lexical search: a close semantic match may not contain the literal string you wanted to cite.
In Archēglyph. Offered alongside BM25. Researchers see both result lists and can switch or combine.
Not to be confused with. RAG layers a generative step on top of semantic search; we stop at the search step and show the retrieved passage directly.