Semantic Chunking
Semantic chunking breaks long documents and web pages into focused, self-contained sections so retrieval systems can pull exact answers without token bloat.
Semantic chunking is the practice of dividing a long document or web page into discrete, logically coherent segments while preserving contextual meaning and structural boundaries.
How Chunking works in practice
Feeding an entire 5,000-word product manual into an LLM to answer a single question wastes context window tokens and introduces irrelevant noise. Chunking isolates individual topics into compact segments (typically 200–500 words).
Effective chunking respects HTML structure—preserving heading hierarchies, lists, and tables intact—and includes overlap between consecutive chunks to prevent cutting thoughts in half.
How SiteMind implements Chunking
SiteMind automatically parses website DOM structures to chunk pages along natural heading hierarchies. You can inspect, edit, or delete individual chunks in your dashboard to instantly update your assistant’s knowledge.
Test our AI tools in your browser (100% Free)
Estimate support savings, token counts, or test prompt injection security guardrails with our zero-cost sandboxes.
Related Technical Concepts
Vector Embeddings
Vector embeddings convert text into high-dimensional numerical lists that capture semantic meaning, enabling computers to compare concepts rather than keywords.
Live Chunk Editing
Live chunk editing allows administrators to modify or delete specific knowledge chunks in the dashboard, instantly updating chatbot answers in under 2 seconds without recrawling.
Context Window Limits
The context window is the maximum number of tokens an LLM can read and process in a single conversation turn, including system instructions, knowledge, and history.
Retrieval-Augmented Generation (RAG)
Retrieval-Augmented Generation (RAG) is a technique where an AI fetches verified passages from a knowledge base before answering, ensuring grounded, factual replies.
Test it on your own website in under 2 minutes.
Enter your domain to index your pages and preview live answers.