A vector database stores numeric embeddings and makes them available for similarity-based retrieval. In semantic search, documents become vectors that preserve relationships in meaning rather than relying only on exact words. section 7.2.2
The underlying idea is distributional similarity: items appearing in similar contexts receive similar embeddings. A search system can therefore retrieve content whose meaning is close to a query, even when the wording differs. section 7.2.2
Similarity in embedding space
Embeddings for apple and banana are closer because they share contexts such as fruit, juice, and eating; car is farther away because its contexts differ. section 7.2.2
Figure 7.6 The process of semantic search
Trace how a query is transformed and matched against stored embeddings during semantic search. section 7.2.2
