AI Glossary Term
What does “Embedding” mean?
Definition
A numeric vector representation that captures semantic meaning of text, images, or other data.
How Embedding works
Example
A help center turns every article into an embedding. When a customer types 'my card was declined', the system embeds the question and finds the articles whose vectors are closest, even if they use different words like 'payment failed'.
Why it matters
Embeddings let software compare meaning with simple math. They underpin semantic search, recommendations, and retrieval for chatbots.
Common questions
How long is an embedding?
Typically hundreds to a few thousand numbers, depending on the model.
Can I compare embeddings from different models?
No. Each model has its own space, so vectors are only comparable if they come from the same model.
Do embeddings capture everything about text?
No. They capture patterns useful for similarity, but can miss details such as exact numbers or negation.