> ## Documentation Index
> Fetch the complete documentation index at: https://docs.corti.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Bias recognition with keyterms
> Configure and manage a custom vocabulary that biases speech recognition toward your domain terms
**Follow along with the runnable example.** This guide walks through the [Dictionary terms example](https://github.com/corticph/corti-examples/tree/main/sdk/typescript/applets/src/applets/dictionary-terms) in `corti-examples` — bootstrap your build from it or use it as reference.
Keyterms bias speech recognition toward a **custom vocabulary** — drug names, proper nouns, acronyms, and other domain terms the model may not otherwise get right. Unlike [replacements](/stt/guides/dictation-replacements), which rewrite the *final* text, keyterms act *before* transcription to make the model more likely to produce the term in the first place.
This guide shows how to configure keyterms through the [Dictation Web Component](/sdk/dictation/overview) and manage a reusable term set.
For how keyterms behave — case-sensitivity, the per-connection and per-term limits, and roadmap — see the [Custom Dictionary Keyterms](/stt/keyterms) reference. This page focuses on wiring them into your app.
***
## Configure keyterms
Add a `keyterms` object to your `dictationConfig`. Its `terms` array holds one `{ term }` entry per word or phrase.
`dictationConfig` is an object, so it must be set as a **JavaScript property** — not an HTML attribute. Rebuild and reassign the config when your term set changes; it re-applies on the next connection.
Keyterms are **case-sensitive**, so define each term with the casing you want preserved (e.g. `"Corti"`, not `"corti"`). Each term is limited to 50 characters, with up to 1,000 terms per connection — see the [reference](/stt/keyterms) for details.
***
## Manage a term set
As with replacements, terms usually come from a catalog plus user-defined entries. Model each with a stable `id` so it can be edited, removed, or protected, then derive the config from the active set.
```ts title="JavaScript" theme={null}
interface Term {
id: string;
term: string;
builtin?: boolean; // protect catalog entries from deletion
}
// A small medical starter catalog.
const CATALOG: Term[] = [
{ id: "t-corti", term: "Corti", builtin: true },
{ id: "t-lisinopril", term: "lisinopril", builtin: true },
{ id: "t-metformin", term: "metformin", builtin: true },
{ id: "t-atorvastatin", term: "atorvastatin", builtin: true },
];
// Derive the config from the active terms.
function buildKeytermsConfig(primaryLanguage: string, items: Term[]) {
return {
primaryLanguage,
keyterms: { terms: items.map(({ term }) => ({ term })) },
};
}
```
Persist a user's term set per **API client** (e.g. keyed by `clientId:tenant`) so their vocabulary follows them across sessions without leaking between tenants.
***
## Export and share a term set
To move a term set between environments or check it into source control, serialize it in the config shape so it round-trips cleanly:
```json title="corti-terms.json" theme={null}
{
"type": "config",
"configuration": {
"keyterms": {
"terms": [
{ "term": "Corti" },
{ "term": "lisinopril" }
]
}
}
}
```
***
## Next steps
Case-sensitivity, limits, and roadmap for the keyterms feature.
Turn spoken command phrases into real editor actions.
The complete, runnable Dictionary terms example in `corti-examples`.
Please [contact us](mailto:help@corti.ai) for help or questions.