Strategies for Splitting Large Documents for LLMs

Processing a 100-page contract with an LLM? You'll need to break it down. But how you break it down affects the quality of your output.

Overlap is Key

When splitting text, it's crucial to have overlap between chunks (e.g., 50 tokens). This ensures that context isn't lost at the boundaries. If a sentence is cut in half, the overlap ensures the full sentence appears in at least one chunk.

Visualizing the Chunks

It's hard to code a splitter blindly. Our Text Chunker visualizes exactly where the splits happen, allowing you to tune your chunk size and overlap parameters interactively.