E · Embedding · evaluation · EU AI Act

Embedding

A numeric vector that represents the meaning of a text fragment.

In one sentence

An embedding is a high-dimensional numeric vector produced by a model that captures the semantic content of a text fragment so that similar meanings end up near each other in vector space.

When it matters

Embeddings are the substrate of semantic search and the bottleneck of recall: changing embedding model = rebuilding the index.

A real-world example

helpcode KB uses 1536-dimensional embeddings; re-indexing a 1,200-page manual takes 4-8 minutes on commodity hardware.

Curated by helpcode research team · Last reviewed 2026-05-22