Glossary

Embeddings

Embeddings are numerical vector representations of objects (text, images, audio) in a multi-dimensional space, where semantically similar objects are positioned close to each other. The foundation of semantic search and recommendation systems.

Example

The phrases "PHP developer" and "backend programmer" will have similar vectors. "Cat" and "dog" are closer to each other than "cat" and "car".

How to obtain

Through an embedding model: OpenAI Embeddings API, Sentence Transformers, BERT, FastText. The model takes text and returns a vector (typically 768–3072 numbers).

Use cases

Semantic search: find documents similar in meaning to a query
RAG: find relevant context for an LLM
Recommendations: "similar products", "related articles"
Text clustering and classification
Duplicate detection