What is the primary goal of contextual retrieval in a RAG pipeline?	To improve retrieval accuracy by adding context to individual text chunks that are otherwise isolated.
What problem does contextual retrieval aim to solve during the chunking process?	That individual text chunks often lose the broader context of the original source document.
What is the pre-processing step involved in contextual retrieval?	Sending an individual text chunk and the overall source document to an LLM (like Claude) to generate surrounding context.
What is a "contextualized chunk"?	The combination of the original text chunk and the additional context generated by the LLM.
What is the main challenge when applying contextual retrieval to a source document?	The original source document may be too large to fit into a single prompt for the LLM.
How can contextual retrieval be applied if the source document is too large?	By including a limited selection of starter chunks (e.g., abstract) and chunks immediately preceding the target chunk.
After generating a contextualized chunk, where is it used in the RAG pipeline?	As the input to the vector index and the BM25 index.
What is the expected outcome of using contextual retrieval on complex documents?	Significantly better accuracy in the RAG pipeline due to richer input context.