Question: 21

You are tasked with developing an AI-powered application using Databricks to summarize long-form legal documents. The documents can be thousands of words long, and the large language model (LLM) has a token limit of 4096 tokens. You need to decide on the optimal chunking strategy to ensure that the summarization captures the essential legal clauses accurately without missing important context. Which chunking strategy is most appropriate to generate an accurate and coherent summary, considering the token limit and the document structure?