Vector Embedding

A numerical representation of text's meaning used by AI systems to match content by semantic similarity rather than exact keyword overlap. Retrieval systems chunk a page (commonly 300-1000 tokens), embed each chunk into vector space, and match it to a query by closeness in that space, which is why a page's internal heading structure functions as a de facto retrieval unit.