Topic
Embeddings
An embedding represents an input as a vector of numbers that can be compared with other vectors.
At a glance
- Representation
- Numeric vector
- Common application
- Similarity-based retrieval
Overview
An embedding represents an input as a vector of numbers that can be compared with other vectors. In text systems, embeddings are often used for similarity-based retrieval, clustering, and related tasks, with the model determining which relationships the representation captures.
Comparing representations
A system embeds a query and candidate items, then compares their vectors with an appropriate similarity measure. Nearby vectors indicate similarity under that model, not a universal measure of truth or quality.
Using embeddings in a system
Input preparation, model choice, and evaluation affect retrieval quality. Store enough source identity alongside vectors to recover the original material and explain what a retrieved result actually represents.
Sources and review
MOOR's explanatory text is supported by the following source links.
- Embeddings — Cohere
- Vector embeddings — OpenAI