What is the primary limitation of semantic search mentioned in the transcript?	It can produce surprising or irrelevant results in corner cases.
What general strategy is proposed to improve RAG search results?	Implementing a separate lexical search system alongside the semantic search.
How is lexical search defined in the context of this pipeline?	It is a classic text search that breaks down the user's question into individual words to find matching text chunks.
What is the specific lexical search method discussed for RAG pipelines?	BM25 (Best Match 25).
What is the first step in the BM25 algorithm?	Tokenizing the user's query (breaking it up into separate chunks).
How does the BM25 algorithm determine the importance of a search term?	By counting its frequency across all documents; infrequent terms are given higher importance.
What is the final step in the BM25 algorithm?	Finding the text chunk that uses the highest weighted terms most often.
What is the ultimate goal of merging semantic and lexical search systems?	To achieve a better balance of search results, combining the semantic aspect with the plain text search aspect.
What two search systems are implemented in the pipeline?	Semantic search (using embeddings and vector databases) and Lexical search (using BM25).
What is the function of the BM25 index store?	To store text chunks and allow for searching based on term frequency and weighted importance.