7 terms
Chunking in retrieval systems splits documents into smaller passages that can be embedded, indexed and retrieved as context for a language model.
Embeddings are dense numeric vectors that encode the meaning of text, images or other data so that similar items sit close together for retrieval.
A knowledge graph represents entities and their typed relationships as a graph with defined semantics, used in AI for search, reasoning and grounded retrieval.
RAG (retrieval-augmented generation) feeds a language model relevant passages retrieved from external sources so its answers rest on current, verifiable data.
Reranking is a second retrieval stage that rescores a shortlist of candidate documents with a stronger model so the most relevant ones reach the top.
Semantic search retrieves documents by meaning rather than exact keywords, comparing vector embeddings of the query and the content in AI retrieval systems.
Vector similarity measures how close two embedding vectors are, using cosine similarity, dot product or Euclidean distance, to find related items.