Embeddings

What it is

An embedding is a numeric vector produced by a model to represent the meaning or features of an item, such as a sentence, product, image, user, or document.

Why it matters

Embeddings make similarity searchable. They power semantic search, recommendations, clustering, retrieval-augmented generation, deduplication, and personalization.

How it works

A model maps inputs into a vector space where similar items are near each other. Systems store vectors and compare them using cosine similarity, dot product, or distance metrics.

When to use it

Use embeddings when keyword matching is too literal or when recommendations should use learned similarity.

Tradeoffs / failure modes

Embeddings can be stale, biased, expensive to recompute, and hard to explain. Model changes often require reindexing.

Tiny example

A document question-answering system embeds uploaded files and retrieves the chunks closest to the user’s question.

Related guides and systems

See Recommendation Engine, Full-Text Search, ChatGPT, YouTube, and Google Photos.