Vector embeddings

Vector embeddings are high-dimensional numerical arrays that convert unstructured data (text, images) into a mathematical format, capturing semantic meaning and relationships for machine learning processing.

We use vector embeddings to transform complex data—like a sentence or an image—into a fixed-size array of floating-point numbers (e.g., 768 or 1536 dimensions). This process maps data points into a high-dimensional space where geometric distance directly correlates to semantic similarity: closer vectors mean more related concepts. Models like BERT or OpenAI’s text-embedding-ada-002 generate these vectors, enabling critical AI functions. Key applications include semantic search, recommendation engines, and Retrieval-Augmented Generation (RAG) systems, allowing algorithms to efficiently process meaning, not just keywords.

https://www.elastic.co/what-is/vector-embeddings