Databricks Certified Generative AI Engineer Associate Practice Questions

Databricks Certified Generative AI Engineer Associate exam preparation

Databricks Certified Generative AI Engineer Associate practice questions in English: 162 questions with explanations. The mock exam has 45 questions and 90 minutes.

Databricks Certified Generative AI Engineer Associate free question 3 / 10 · Domain 2: Data Preparation

A team splits its internal work-procedure manuals (plain text) into chunks for RAG. Each paragraph of a manual describes one task, so the team wants paragraph breaks to be the chunk boundaries and keep each block of content intact. Which LangChain splitter does the Databricks AI Cookbook article on the unstructured data pipeline give as an example of this approach?

  1. MarkdownHeaderTextSplitter
  2. SemanticChunker
  3. CharacterTextSplitter
  4. RecursiveCharacterTextSplitter
Show the answer and explanation

Answer: D

Why this is correct
The Cookbook divides chunking into four approaches. For paragraph-based chunking, which decides chunk boundaries at the natural paragraph breaks of the text, it gives LangChain's RecursiveCharacterTextSplitter as an example. Paragraphs often hold related information together, so splitting at paragraph boundaries makes it easier to keep the semantic coherence of the chunks. That is the reason to use this approach. For a manual whose paragraphs are separated by task, one chunk is likely to hold one task.

Why the other options are wrong
"CharacterTextSplitter" … The Cookbook uses CharacterTextSplitter as an example of fixed-size chunking, which splits at a set number of characters or tokens. It is easy to put together, but it rarely produces chunks with a coherent meaning, and the Cookbook says it is rarely usable in production.
"MarkdownHeaderTextSplitter" … MarkdownHeaderTextSplitter is an example of format-specific chunking, which uses the structure of the format itself, such as Markdown headings, as boundaries. A plain-text manual without heading markup has no such cue for boundaries.
"SemanticChunker" … The Cookbook names SemanticChunker as an example of semantic chunking, which analyzes the content with techniques such as topic modeling and places boundaries where the topic shifts. It infers boundaries from the content instead of using visible paragraph breaks, which takes more effort than the basic approaches.

Key point
The Cookbook says the best chunking approach and size depend on the use case and the nature of the data, that no single method suits every case, and that you should decide by experimenting. When choosing, also decide the chunk size (small chunks focus on detail but lose surrounding context; large chunks carry context but mix in unrelated information) and the overlap between neighboring chunks.

Go to the Databricks Certified Generative AI Engineer Associate practice app (10 free questions; purchase unlocks all questions)

See the full practice set