Embedding
An embedding turns a piece of text into a list of numbers positioned so that similar meanings sit close together.
Why it matters
Keyword search fails when the asker and the author use different words for the same thing — 'extra login' and 'additional seat'. Embeddings solve that by comparing meaning rather than characters.
- Search returning nothing for a question the docs answer
- Customers using vocabulary your content never uses
- Synonyms and paraphrases missed entirely by keyword matching
How it works
A model converts text into a vector — typically several hundred to a few thousand numbers. Texts with similar meanings produce vectors that are close together, so finding relevant passages becomes a matter of measuring distance rather than matching strings.
Where it fits in retrieval
Your content is chunked, each chunk is embedded, and the vectors are stored. A question is embedded the same way, and the nearest stored vectors are the passages most likely to be relevant.
Index time
Every chunk is embedded once and stored.
Query time
The question is embedded and compared against the store.
Its limits
Embeddings capture semantic similarity, which is not the same as correctness. Two passages can be about the same topic while only one answers the question, which is why retrieval systems often re-rank results rather than trusting distance alone.
Common questions
- Do embeddings understand meaning?
- They encode statistical relationships from training in a way that behaves remarkably like meaning for retrieval. Whether that constitutes understanding is a philosophical question; for building search, the behaviour is what counts.
- Are embeddings the same as keywords?
- No. Keywords match characters, embeddings match meaning, which is why an embedding-based search finds the seats article when someone asks about extra logins.
From the blog
All posts- MeasuringDeflection rate is the most overstated number in support automationCounting bot conversations as deflected tickets overstates the result. The honest version is a before-and-after on ticket volume, read next to satisfaction.Read
- EvaluatingHow to test an AI support tool before you trust itEvery AI support tool demos well, because demos ask questions the content covers. Four questions that separate them, and what a good answer looks like.Read
- MeasuringResolution rate vs deflection: stop mixing the twoResolution and deflection answer different questions. Mixing them inflates the result and hides whether customers actually got what they needed.Read
Answer honestly. Capture the rest.
Point Matter Chat at your site and see what it can — and can't — answer. It's honest about both.
No credit card. 2 minute setup.
Every answer cites the source it came from. When there isn't one, it says so — and hands the visitor to you.
Installs on the tools you already run.