Does king minus man plus woman really equal queen?

It works often enough to be a famous demonstration and not reliably enough to be a rule. Queen is usually the nearest neighbour, partly because the search normally excludes the input words themselves. What survives the caveats is the real point: directions in embedding space carry meaning.

Is word2vec still used?

Rarely on its own. Modern models give every occurrence of a word its own context-dependent vector, so bank in a river sentence and in a money sentence get different positions, which word2vec could not do. The idea it proved, meaning as geometry, is still the foundation.

What other relationships show up as directions?

Singular to plural, present to past, country to capital, and many more, each as a roughly consistent offset. They are learned regularities in how words co-occur, not stored facts, so they hold up on common cases and break on rare ones.

Plain-language explainer

Word vector analogies, explained

Why does king minus man plus woman land near queen?

Because when words become vectors, relationships become directions. In a well-trained embedding space, the arrow from man to woman points roughly the same way as the arrow from king to queen: a 'gender direction' the model learned purely from how words co-occur in text. Add that arrow to king and you land near queen. The famous demo, from word2vec in 2013, was the first striking proof that meaning could be geometry, the idea every modern model is built on.

Last reviewed July 25, 2026

Do not just read it. Operate the mechanism yourself in a short interactive lesson.

See it work: Embeddings: meaning as coordinates →

Free, no code, no signup.

What people get wrong

It is just a party trick. The property behind it, relations as directions, is exactly why semantic search and recommendations work.
Chat models answer analogies by doing this arithmetic. It is a property of the embedding space, not the procedure the model runs.
The math is exact. Analogies are approximate and the showcased examples are the ones that land; plenty of others miss.

Where you see it in real products

Search engines match 'cheap flights' to 'low-cost airfare' because both live in the same region of the space.
'Similar items' recommendations are nearest neighbors in an embedding space.
Every ML course opens with this demo because it makes vectors mean something.

Common questions

Does king minus man plus woman really equal queen?: It works often enough to be a famous demonstration and not reliably enough to be a rule. Queen is usually the nearest neighbour, partly because the search normally excludes the input words themselves. What survives the caveats is the real point: directions in embedding space carry meaning.
Is word2vec still used?: Rarely on its own. Modern models give every occurrence of a word its own context-dependent vector, so bank in a river sentence and in a money sentence get different positions, which word2vec could not do. The idea it proved, meaning as geometry, is still the foundation.
What other relationships show up as directions?: Singular to plural, present to past, country to capital, and many more, each as a roughly consistent offset. They are learned regularities in how words co-occur, not stored facts, so they hold up on common cases and break on rare ones.

Related explainers

Embeddings Vector databases Transformers

One idea at a time, in your inbox

New lessons and explainers, written the way these pages are. Now and then, not daily, and never a sales sequence.

We email you a confirmation link first. Unsubscribe in one click, any time. Privacy.

Part of See How AI Works, a free interactive course, where you learn how modern AI works by operating it, not watching videos.