AI Glossary

Chunking

What does “Chunking” mean?

Chunking: Splitting long documents into smaller pieces before creating embeddings, so retrieval can return just the relevant passage. Chunk size and boundaries matter: pieces that are too small lose context, while ones that are too large dilute relevance. Splitting by headings or paragraphs often works better than fixed character counts.

Chunking visual guide: Splitting long documents into smaller pieces before creating embeddings, so retrieval can return just the relevant passage. Chunk size and boundaries matter: pieces that are too small lose context, while ones that are too large dilute relevance. Splitting by headings or paragraphs often works better than fixed character counts.
Visual guide Open full size (new tab)

What is an example of “Chunking”?

A 200-page manual might be chunked by section, so a question about warranty terms retrieves only the warranty section.

Keep learning

See also

Retrieval-augmented generationTermEmbeddingsTermVector databaseTerm

Same topic

Context windowTermChain of thoughtTermBatch APITermFew-shot promptingTermAPITerm