What are the four basic steps in a typical RAG pipeline?	Take a source document, break it into chunks, find relevant chunks, and put them into a prompt.
Why is the process of chunking a critical step in the RAG pipeline?	How the document is chunked has a huge impact on the quality and accuracy of the entire RAG pipeline.
Define size-based chunking.	Dividing a large document into strings of text that are of more or less equal length.
What are the two primary drawbacks of simple size-based chunking?	Chunks often contain cutoff words, and they tend to lack surrounding context.
How does the overlap strategy improve size-based chunking?	It includes a portion of text from neighboring chunks, providing the current chunk with more context.
Define structure-based chunking.	Dividing text based on the overall structure of the document, such as headers, paragraphs, or sections.
What is a major limitation of structure-based chunking?	It requires predictable document formatting (like Markdown) and may fail on plain text documents (e.g., PDFs).
Define semantic-based chunking.	Using Natural Language Processing (NLP) to group consecutive sentences or sections based on how related they are.
Which chunking strategy is described as the most advanced?	Semantic-based chunking.
What are the default chunk settings for chunking?	A chunk length of 150 with an overlap of 20.
What is the primary drawback of using the default chunk settings?	They often do not provide enough meaning or context within the chunks.
What is the typical configuration of the "chunk by sentence" strategy?	Five sentences per chunk with one sentence of overlap, using a regular expression to split.
What is the main advantage of using the "chunk by section" strategy?	It ensures that each chunk contains exactly one structural section (e.g., executive summary, section one).
What factor determines which chunking strategy should be used?	The nature and guaranteed structure of the document being processed.
When is the "chunk by section" strategy unlikely to work effectively?	When the document lacks formatting (e.g., user-provided text).
If sentence or section chunking fails (e.g., when chunking code), what is the final fallback strategy?	Chunk by character.
How reliable is the "chunk by character" strategy?	It is not guaranteed to give the best results, but it works in the vast majority of cases.