Chunking
Splitting documents into smaller passages before they are embedded and indexed for retrieval. How documents are chunked strongly affects whether a retrieval system finds the right evidence.
Why it matters
Chunks that are too small lose context, such as the heading that explains what a table means. Chunks that are too large dilute relevance and consume the context window. Poor chunking is a common cause of weak retrieval-augmented generation.
Good chunking respects document structure, such as sections, clauses and tables, and carries metadata like the source, date and permissions with each chunk.
In practice
For example, a UK housing regulator’s guidance assistant might chunk policy documents by numbered section, prefix each chunk with its section title and publication date, and keep tables whole, so answers cite the correct paragraph of current guidance.
Where Rodan fits
Rodan designs ingestion, chunking and evaluation for retrieval systems in AI and Decision Systems work. See also what is RAG and why does it matter for enterprise AI.