Day 3 - Chunking - RAG

Short summary

Chunking breaks documents into segments for vector embeddings, enabling RAG systems to retrieve specific answers rather than full passages. Fixed chunking loses semantic boundaries; overlapping chunks and sentence-based approaches preserve meaning and reduce vector distance between related content.

• •Chunking solves the problem where entire passages embedded as single vectors return full documents instead of relevant excerpts
• •Fixed character-based chunking loses meaning at boundaries; sentence-aware chunking preserves semantics
• •Overlapping chunks reduce vector distance between related segments, improving retrieval precision

Generated with AI, which can make mistakes.