Design Retrieval Systems Using Advanced Chunking Strategies GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   In fixed-size chunking, what is the purpose of the 'overlap' parameter?

   • ATo make the chunk size variable so that each chunk ends at a sentence boundary.
   • BTo reduce the total number of chunks generated from a document.
   • CTo ensure that no text is lost when splitting a document into chunks.
   • DTo allow chunks to share some text so that context is not abruptly cut off at boundaries.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Why is chunking necessary in a retrieval system that uses embeddings to find relevant text passages?

   • ATo eliminate the need for a vector database by storing text directly in memory.
   • BTo ensure each text segment is small enough for the embedding model to process and to improve the precision of similarity search.
   • CTo make the retrieval system faster by reducing the total number of documents in the corpus.
   • DTo reduce the number of tokens sent to the embedding model, making the embedding process faster.
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   A team is building a RAG system over a large corpus of technical documentation. They have a strict budget for embedding costs, so they want to minimize the number of chunks. However, they also need to ensure that each chunk retains enough context to answer follow-up questions. They are considering using a sliding window with overlap. Which trade-off should they consider?

   • AOverlap has no effect on embedding cost.
   • BIncreasing overlap improves context retention but increases the number of chunks and embedding cost.
   • CIncreasing overlap reduces the number of chunks and embedding cost.
   • DDecreasing overlap improves context retention and reduces cost.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A customer support team is building a RAG system over product manuals. The manuals are written in plain text with paragraphs and bullet points. The team tried fixed-size chunking with 400 tokens and found that chunks often split bullet lists, making the retrieved content hard to read. They want chunks that respect the natural flow of the text while keeping a target size. Which approach should they use?

   • AFixed-size chunking with a smaller chunk size of 200 tokens.
   • BSemantic chunking using sentence embeddings.
   • CRecursive character text splitting with separators for double newlines, newlines, and periods.
   • DDocument-based chunking by page breaks.
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A team is building a RAG system over a long document that is a single continuous narrative, like a novel. They want to ensure that each chunk retains enough context to answer questions about the plot, but they also want to minimize the number of chunks to reduce embedding costs. Which approach should they use?

   • AUse recursive character text splitting with no overlap.
   • BUse document-based chunking by chapters.
   • CUse semantic chunking to group related plot points.
   • DUse fixed-size chunking with a moderate chunk size and a small overlap to retain some context while keeping the chunk count low.
   Reveal Solution Discussion Select an answer first