Learning path · Retrieval & Ranking · 43

Semantic Search

Finding documents by meaning similarity between query and corpus embeddings rather than exact keyword match.

Why it matters

Captures paraphrases and conceptual questions keywords miss.
Core retrieval stage before RAG generation.
Fails on rare proper nouns without hybrid lexical backup.

Key ideas

Query embedding
Top-K retrieval
Similarity thresholds

Top resources

01DocsPinecone
Understanding embeddings
Why this resource. The retrieval primitive under every “search by meaning” feature.
Covers in this concept
- dense retrieval
- kNN
02DocsPinecone
Hybrid search
Why this resource. When meaning-only search fails on SKUs and IDs.
Covers in this concept
- lexical gap

Semantic search embeds the question, pulls nearest neighbours, and hands them to a reranker or the LLM. Too few hits miss evidence; too many waste tokens. Log zero-hit queries: they usually mean a stale index or a missing glossary term. Filter by ACL on every query so embeddings never leak another tenant's docs.

Updated 2026-08-09 · Full learning path