Category:
Category:
Category:
Retrieval & RAG
Definition
The process of breaking documents into smaller segments for retrieval.
Explanation
Chunking ensures that retrieved text segments are meaningful and relevant. Poorly chosen chunk sizes lead to irrelevant or noisy retrieval. Methods include fixed-size chunking, sliding window chunking, semantic chunking, and hybrid chunking. Chunking influences RAG accuracy, retrieval relevance, LLM grounding, and hallucination mitigation.
Technical Architecture
Document → Chunker → Embedding → Vector Database → Retrieval → LLM
Core Component
Chunker, embedding model, overlap window, metadata tags
Use Cases
RAG, search engines, knowledge assistants, research tools
Pitfalls
Chunks too large lose relevance; too small lose meaning; improper boundaries degrade retrieval quality
LLM Keywords
RAG Chunking, Semantic Chunking, Ddocument Segmentation
Related Concepts
Related Frameworks
• RAG
• Embeddings
• Reranking
• Vector DB
• Chunking Decision Framework