The Semantic Search Guide
Plain explanations of how search by meaning works, from embeddings and hybrid search to evaluating and building it.
Try: How to build semantic search What is semantic caching What are embeddings
184guides
18topic groups
27hands-on tutorials
30comparisons
Semantic search finds results by meaning, not exact words. What it is, how embeddings make it work, and when it helps or fails against keyword search.
Read the guidePopular guides
The topics people search for most.
- BasicsSemantic search in AI, NLP and machine learning
- How it worksSimilarity metrics: cosine, dot product, Euclidean
- EmbeddingsRe-embedding when the model changes
- How it worksSemantic indexing and vector indexes
- Search evaluationRecall@k
- Hybrid searchWhat is hybrid search
- How it worksApproximate nearest neighbor search
- How it worksChunking for semantic search
Every guide, by section
All 184 guides in 18 groups.
Basics7
What semantic search is, with examples and limits.
The pipeline from text to ranked results.
- How semantic search works
- Semantic search architecture
- Vector search
- Similarity search
- Similarity metrics: cosine, dot product, Euclidean
- Dense vectors
- Sparse vectors
- Dense vs sparse vectors
- Approximate nearest neighbor search
- Exact vs approximate search
- Semantic indexing and vector indexes
- Top-k retrieval
- Relevance scores
- Synonyms and related terms
- Multilingual and cross-lingual semantic search
- Semantic matching
- Chunking for semantic search
The vectors that carry meaning, and how to choose them.
- What are embeddings
- Choosing an embedding model
- MTEB leaderboard
- Domain-specific embeddings
- Fine-tuning embedding models
- Embedding dimensions
- Re-embedding when the model changes
- Batch embedding
- Asymmetric vs symmetric semantic search
- Embedding and vector quantization
- Matryoshka embeddings
- Multimodal embeddings
- Tokenization vs embedding
- Word vs sentence embeddings
BM25 and the lexical search semantic search builds on.
- What is keyword search
- BM25 algorithm
- TF-IDF
- Inverted index
- Full-text search
- Vocabulary mismatch problem
- Stemming, lemmatization and analyzers
Semantic search set against its alternatives.
- Semantic search vs keyword search
- Semantic search vs vector search
- BM25 vs vector search
- Semantic search vs similarity search
- Natural language search
- Semantic search vs RAG
- Semantic search vs knowledge graph search
- Semantic search vs the semantic web
- Search engine vs vector database for semantic search
- pgvector vs a dedicated vector database
Keyword and vector search run together.
- What is hybrid search
- How hybrid search works
- When to use hybrid search
- Hybrid search vs vector search
- Reciprocal rank fusion
- Relative score fusion
- Score fusion methods
- Balancing keyword and vector scores
- Hybrid search for RAG
- Hybrid search in SQL databases
- Databases that support hybrid search
- Hybrid search on distributed databases
- Hybrid search tutorial
- Hybrid search for enterprise search
- Score normalization in hybrid search
Dense, sparse and late-interaction retrievers.
- Retrieval models explained
- Dense retrieval
- Sparse retrieval
- Dense vs sparse retrieval
- SPLADE and learned sparse retrieval
- Bi-encoders
- Cross-encoders
- Bi-encoder vs cross-encoder
- ColBERT and late interaction
- Multi-vector search
- Agentic retrieval and agentic search
Reordering the top results with stronger models.
- What is reranking
- Cross-encoder rerankers
- Best reranking models
- LLM reranking
- Reranking in RAG
- Maximal marginal relevance
- Reranking latency and cost
What happens to a query before retrieval.
- Query understanding
- Query expansion
- Query rewriting
- HyDE (hypothetical document embeddings)
- Multi-query retrieval
- Query decomposition
- Query intent classification
- Typo tolerance and spelling correction
- Semantic autocomplete
Combining meaning with structured filters.
- Metadata filtering
- Pre-filtering vs post-filtering
- Filtered ANN search
- Filter selectivity
- Vector databases with metadata filtering
- Searching vs filtering
- Bitmap filters
- Faceted semantic search
- Permission-aware search
- Multi-tenant semantic search
Measuring whether search actually works.
- How to evaluate semantic search
- Information retrieval metrics
- Recall@k
- Precision@k
- Mean reciprocal rank
- nDCG
- Relevance vs recall
- ANN recall
- Search test sets and relevance judgments
- Retrieval benchmarks
- Online search metrics and analytics
- Why vector search returns irrelevant results
- Monitoring search quality in production
Latency, cost and freshness at scale.
- Scaling semantic search
- Search latency
- Cost of semantic search
- Real-time indexing and index freshness
- Managed vs self-hosted search
- GPU acceleration for vector search
- Recall vs latency trade-off in semantic search
Reusing answers for questions that mean the same.
- What is semantic caching
- Semantic caching for LLM apps
- GPTCache
- Similarity thresholds for caching
- Semantic cache invalidation
- AI gateway semantic caching
- Semantic caching vs prompt caching
Tutorials from Python to Postgres and Elasticsearch.
- How to build semantic search
- Semantic search with Python
- Semantic search with Postgres and pgvector
- Semantic search with Elasticsearch
- Semantic search with OpenSearch
- Choosing a vector database for semantic search
- Semantic search with LangChain and LlamaIndex
- Semantic search over documents and PDFs
- Designing a search index schema
- Migrating semantic search to a new vector database
- Semantic search with Pinecone
- Semantic search with Weaviate
- Semantic search with Qdrant
- Semantic search with Milvus
- Semantic search with Chroma
- Semantic search with Redis
Engines, platforms and APIs compared.
- Semantic search tools compared
- Open-source semantic search engines
- Semantic site search tools
- Enterprise semantic search platforms
- Semantic search APIs
Interactive tools for cost, storage, similarity and ranking.
- Embedding cost calculator
- Vector storage and memory calculator
- Vector similarity calculator
- Reciprocal rank fusion calculator
- Chunk size calculator
Where semantic search pays off.
- Semantic search use cases
- Semantic product search for ecommerce
- Semantic site search
- Enterprise and internal document search
- Customer support and help center search
- Semantic recommendations
- Semantic search in RAG
- Image, video and audio search
- B2B semantic search
- Patent and legal search
- Semantic code search
- Academic and research paper search
The information retrieval theory underneath.
- What is information retrieval
- Vector space model
- Latent semantic indexing
- Word2vec and GloVe
- Transformers and BERT in search
- Sentence-BERT
- High-dimensional vector space
- Curse of dimensionality
- Semantic similarity vs relatedness
- Semantic search at Google
- Knowledge graphs in search
- Cache hits, misses, cold and warm caches