What the embedding explorer shows
Explore word embeddings laid out in 2D: hover for labels, search to filter, and try vector arithmetic like king − man + woman ≈ queen. Words with similar meanings cluster together, so you can literally see the geometry of language a model learns.
How embeddings work
An embedding maps each word (or sentence, or image) to a list of numbers — a vector — such that distance encodes meaning: related concepts land near each other, and directions in the space capture relationships (gender, tense, plurality). Real embeddings have hundreds of dimensions; this view projects them down to two so the structure is visible, which is why the famous analogies show up as simple vector math.
Why it matters
Embeddings are the backbone of modern AI search and retrieval. Semantic search, RAG (retrieval-augmented generation), recommendations, and clustering all work by comparing vectors instead of matching keywords. Learn the terms in the glossary: vector embeddings and semantic search.
Frequently asked questions
What is an embedding, in one sentence?
A numeric vector representing a piece of content, positioned so that similar meanings are close together in the space.
Why does king − man + woman ≈ queen work?
Because the model learns consistent directions for concepts like gender, subtracting and adding those directions moves you between related words.
How does semantic search use embeddings?
It embeds your query and every document, then returns the documents whose vectors are nearest — so it matches meaning, not just exact words.