Data & retrieval
Chunking
Splitting documents into smaller units for indexing and retrieval.
Definition
Chunking divides source material by length, structure or meaning before creating embeddings or passing context to a model.
Why it matters
Poor boundaries can separate a financial figure from its qualifier, footnote or governing policy and produce misleading answers.
Related concepts
- Embedding
A numerical vector representing the meaning or features of an item.
- Citation correctness
Whether a cited source genuinely supports the associated claim.
- Retrieval-augmented generation
Generating answers using information retrieved at request time.