5 exercises on vector embeddings search vocabulary.
0 / 5 completed
1 / 5
What is a vector embedding?
Embedding: a model maps content to a dense vector of numbers so that similar items lie close together in the vector space. This enables semantic comparison beyond exact keyword matching.
2 / 5
What does cosine similarity measure between two embeddings?
Cosine similarity: compares the direction of two vectors regardless of magnitude. A value near 1 means the items are semantically very similar; near 0 means unrelated. It is the common metric for embedding search.
3 / 5
What is approximate nearest neighbor (ANN) search?
ANN: exhaustively comparing a query against millions of vectors is slow. ANN indexes like HNSW return the likely nearest neighbors much faster, accepting a small chance of missing the absolute closest match.
4 / 5
What does semantic search achieve over keyword search?
Semantic search: by comparing embeddings rather than literal tokens, it retrieves conceptually related results. A query for "how to reset password" finds a doc titled "account recovery steps" despite no shared keywords.
5 / 5
What is dimensionality in the context of embeddings?
Dimensionality: the length of the embedding vector (e.g., 768 or 1536). Higher dimensions can capture more nuance but cost more memory and compute. Models output a fixed dimensionality you must store consistently.