A retrieval system using learned embeddings to find semantically similar documents via vector similarity.