Split a procedure manual into RAG retrieval units.
Keep prerequisites across paragraph boundaries.
Choose two evaluation practices.
Preserve meaningful boundaries, compare overlap amounts, and measure recall, evidence quality, duplicate results, and cost.
Detailed explanation
Procedure and condition relationships can be lost.
Context remains available while duplicate retrieval can be measured.
The split's quality and operational trade-offs become visible.
Candidates, duplicate answers, and cost may grow.
Check official documentation
Compare the following points with the official documentation.
Amazon Bedrock Knowledge Bases公式のデータ分割とAIF-C01 Domain 3.2を確認する。
Official references
What to confirm
チャンクサイズと重複を検索品質・根拠・費用のトレードオフとして評価できる。
Key points
- Meaningful boundaries
- Overlap
- Recall
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
Chunk
A bounded document fragment used for embedding, retrieval, or context assembly.
Overlap
Content repeated between adjacent chunks to reduce loss at a split boundary.