You are building a retrieval system over mixed document types, including policy pages, long technical docs, FAQs, and scanned forms. The team has noticed that one chunk size does not work well across all of them, and answer quality drops when the same chunking rule is applied everywhere.
How do you chunk documents when the right chunk size differs by document type?
You are building a retrieval system over mixed document types, including policy pages, long technical docs, FAQs, and scanned forms. The team has noticed that one chunk size does not work well across all of them, and answer quality drops when the same chunking rule is applied everywhere.
How do you chunk documents when the right chunk size differs by document type?