AI glossary · Data and knowledge
Embeddings
In one sentence
An embedding is a list of numbers that represents the meaning of a piece of text, an image or audio. Items with similar meaning have embeddings that sit close together, which lets you search by meaning rather than by exact words.
Also known as: vector embeddings · text embeddings · semantic vectors · vector representation
What are embeddings?
A computer doesn’t naturally see that “how long does delivery take?” and “when will my order arrive?” are the same question. The words don’t match. Embeddings solve that: a model turns each piece of text into a list of hundreds or thousands of numbers, placed in a space where similar things end up near each other.
Put it to work
UTIO's AI, through an API
From €29/mo + VAT
For developers and agencies building their own products.
See the AI APIThink of a map. Instead of two coordinates it has hundreds, but the idea is the same: “delivery”, “shipping” and “arrival time” land in the same area; “invoice” lands far away. To tell whether two texts are about the same thing, you just measure the distance between their points.
Related to this term
AI API
- Transcription, voice, text, image, OCR and embeddings
- Per-client keys with a spending cap
- Prepaid wallet: €50 minimum top-up
From €29/mo + VAT Panel account, plus usage
See the AI APIHow it works
- Calculate. An embedding model reads the text, split into tokens, and returns its vector: the list of numbers.
- Store. The vectors for your documents are saved, usually in a vector database.
- Compare. When a question comes in, its vector is calculated and the closest ones are found.
That makes several things possible:
- Semantic search: finding the paragraph that answers the question even if it uses different words.
- RAG: passing those paragraphs to the language model so it answers with your data.
- Classifying and grouping: clustering similar messages and spotting recurring topics.
- Finding duplicates in catalogues or databases.
One caveat: embeddings capture similarity of meaning, not truth. Two contradictory sentences on the same topic can sit very close together.
Example
An online hardware shop in Norwich stocks 8,000 products, and its search only finds exact matches. Anyone typing “hang a picture without drilling” gets nothing, because no product is called that.
The shop calculates an embedding for every product page and every search. Now that search returns adhesive strips, wall hooks and nail-free hangers. The product pages haven’t changed; the way of comparing them has.
Embeddings at UTIO
UTIO’s agents answer from your knowledge base: they search your documentation for the useful passages before replying. You don’t need to calculate embeddings or set anything up for that.
If you’re a developer or an agency, UTIO’s AI API includes an embeddings service: vectors to index and search your own content, with the same key you use for transcription, OCR or text generation. You choose where to store them and how to search them.
Frequently asked questions
What's the difference between a token and an embedding?
A token is a chunk of text, the unit a model reads. An embedding is a list of numbers that sums up the meaning of a whole passage, sentence or extract. The text is split into tokens first, then the model calculates the embedding.
Are embeddings from different models compatible?
No. Each embedding model has its own space of numbers. If you switch models, you have to recalculate the embeddings for all your documents.
Do embeddings work across languages?
Multilingual models place sentences with the same meaning close together even when they're in different languages. So a question asked in English can find a document written in Spanish.
Related terms
- Vector database
A vector database stores embeddings, lists of numbers that represent the meaning of content, and finds the ones closest to a query very quickly. It's the search component in most RAG systems.
- RAG
RAG (retrieval-augmented generation) is a technique in which a language model first retrieves the relevant passages from your own documents and then writes its answer from them, instead of relying on what it memorised in training.
- Knowledge base
A knowledge base is a business's organised store of information (FAQs, terms, opening hours, product sheets, procedures) that an AI agent or a team draws on for answers.
- Tokenization
Tokenization is the process of splitting text into tokens, the pieces a language model understands, and mapping each one to a number in its vocabulary. It's the first step before a model can read or write anything.
- Text classification
Text classification is the AI task of automatically assigning one or more categories to a piece of text, such as an email or a message: type of request, topic, priority or language. It's the basis for sorting, prioritising and routing work.
Updated on