A vector space represents words, phrases, or documents as mathematical vectors: ordered numbers that machine-learning systems can compare. In embedding-based applications, nearby vectors indicate stronger semantic relatedness, while distant vectors indicate weaker relatedness. section 2.8.4 section 10.4.2
Embedding workflows use vector spaces to:
- Represent contextual meaning rather than only dictionary definitions. section 10.4.2
- Compare text chunks through distances or similarity scores. section 2.8.4
- Provide numerical inputs for text classification, sentiment analysis, machine translation, search, and code-related tasks. section 2.8.4
- Support preprocessing and retrieval architectures when text is chunked, embedded, and managed as vectors. section 10.4.2
Simple embedding representation
A toy representation might encode “cat” as [0.2, 0.3, -0.1] and “dog” as [0.4, 0.1, 0.2]. Their positions can then be compared to estimate relatedness. In practice, an embedding model generates vectors from context, and an OpenAI API call returns the resulting values for application use. section 2.8.4
