Chunking Strategies and Embedding Selection

This lesson is for enrolled learners

The concept Chunking is how you cut source documents into the units you index and retrieve. Too large, and a hit contains mostly unrelated text, wasting the prompt and confusing the model. Too small, and a sentence loses the heading, the effective date, or the exception in the next paragraph. There is no universal token length. The right chunk is……

Enroll to continue Already enrolled? Sign in