🤖 AI EngineeringID:
sentence-window-chunkingSentence-Window Chunking
Sentence-Window 句窗口切分
🎯Core Definition
Sentence-Window Retrieval (Small-to-Big Sentence Window) is an advanced RAG chunking architecture that cleanly decouples the 'Retrieval Unit' from the 'Generation Context Unit'; during indexing, documents are parsed into standalone single sentences (or tiny 2-3 sentence chunks) and embedded into the vector store; metadata retains pointers to the surrounding context window (e.g., surrounding 3 preceding and 3 succeeding sentences); upon retrieving the high-precision sentence match, the system dynamically expands and loads the wider sentence window into LLM context.
💡Use Cases
Domain-specific question-answering requiring pinpoint factual retrieval combined with comprehensive surrounding paragraph context for generation.
⚡Key Problems Solved
Large chunks (1024 tokens) dilute embedding representations across broad background noise, reducing cosine similarity for precise queries; sentence-level indexing ensures maximum similarity specificity, while surrounding window expansion provides rich context for coherent generation.
🎯5 High-Frequency Exam Points
1
How does Sentence-Window resolve the fundamental contradiction between embedding information dilution and generation context deficiency?
2
When retrieval hits multiple adjacent sentences in the same paragraph, how to execute window merging to prevent duplicate context injection?
3
Analyze the trade-offs of `window_size` parameter (preceding/succeeding sentences) on LLM token budget vs answer completeness?
4
Quantify the vector count inflation and storage overhead of Sentence-Window compared to traditional overlapping paragraph chunks?
5
How to decouple architecture by offloading expanded window texts to Redis/SQL while vector DB only indexes single sentences?
📖 In-depth Guide:📄 naive-and-advanced-rag →
Updated 2026-08-14🎯
🚀 Start Card Practice ➔Test Your Knowledge: Practice Questions for "Sentence-Window Chunking"
Single choice pitfall questions with instant feedback and mistake tracking.