A Jargon Library is a list of your company's own vocabulary — product names, acronyms, internal jargon, and people's names — that Oak uses when transcribing, so those terms come out right instead of being flattened into common words. You apply a library per project, and it's owned in your own workspace, so your private terms stay yours.
Every company has its own private language. Product names, internal acronyms, the way you talk about clients, the executives whose names keep getting mistranscribed. A Jargon Library is the structured place where this vocabulary lives — and where Oak picks it up, so your team’s terms come out consistently right across every transcript, in every language.
Key takeaways
- It captures your private language. The names, acronyms, and terms that are obvious to your team but invisible to a general model.
- Seed the obvious first. Company name, top products, the people on your calls, and your most-used acronyms catch the majority of mistranscriptions on day one.
- Grow it from corrections. Every time a reviewer fixes a misrendered term, add it once — and it stays right from then on.
- It becomes self-curating. Over a quarter of adding terms as they're missed, the library covers your real vocabulary with little ongoing effort.
- Applied per project, owned in your workspace. A library is scoped to the project it belongs to, while ownership and privacy stay with your workspace — so confidential terms stay yours, which is what makes it safe for legal and regulated teams.
What is a Jargon Library?
A Jargon Library is a per-project list of the words a general transcription model is most likely to get wrong because they’re specific to your organisation. You apply a library to the project it belongs to, and it lives in your own workspace — so a project’s vocabulary stays scoped to that project, while ownership and privacy stay with you. A generic model has never heard your product’s codename, your CEO’s name, or your industry’s in-house abbreviation, so when one comes up it substitutes the nearest ordinary word that sounds similar. The Jargon Library closes that gap: it tells the model, in advance, the vocabulary that project’s meetings actually use, so those terms are recognised rather than guessed. Combined with the engine’s context-based decoding, it resolves most of your company-specific ambiguities on the first pass — the difference between a transcript you can use and one you have to clean by hand.
What goes in it
Five categories cover almost everything worth adding:
| Category | Examples |
|---|---|
| Names | People, products, and programmes — anything proper-noun that recurs on your calls |
| Acronyms | Your internal abbreviations, and their expansions |
| Industry terms | Sector-specific vocabulary a general model wasn't trained on |
| Internal jargon | Phrases your team uses that nobody outside the company would |
| Client-specific terms | Vocabulary unique to a particular customer or matter |
How to seed it
You don’t need to be exhaustive on day one — you need to cover the terms that recur. Start with roughly the obvious fifty entries:
- Your company name and any sub-brands or programme names.
- Your top product names — the handful that come up on almost every call.
- The people who appear on calls — the executives and team members whose names recur.
- Your most-used acronyms — the internal shorthand that fills your meetings.
That first pass catches the majority of mistranscriptions immediately, because a small number of terms account for most of the errors. Everything else can be added as it comes up.
How to grow it
The library grows almost by itself once the habit is in place: every time a reviewer corrects a misrendered term, add it to the library. Because the same vocabulary recurs across a team’s meetings, each addition pays off many times over, and within a quarter the library becomes effectively self-curating — it covers your real working vocabulary without anyone having to sit down and brainstorm it. Libraries can grow large as a company’s vocabulary is captured, and a well-curated one keeps the transcripts consistent meeting after meeting. The same discipline is described from the transcription side in Cantonese business jargon.
Per project, owned in your workspace
A Jargon Library works along two dimensions. It applies per project: you attach a library to the project it belongs to, so a matter’s or a show’s vocabulary is used where it’s relevant and doesn’t spill into unrelated work. And it is owned in your own workspace: the library and the terms in it belong to you, so a law firm’s privileged terms, a media company’s unreleased show names, and a client’s confidential vocabulary all stay within your control and never become anyone else’s. That combination — scoped to the project for accuracy, owned in your workspace for privacy — is what makes the feature usable for Legal and Statutory deployments, where the vocabulary itself can be sensitive. The library makes your transcripts more accurate precisely because it knows your private terms, and the ownership model is what ensures that knowledge stays yours.
Where this shows up
Across every Oak deployment, but most visibly in Legal, where preserving precise terminology is the whole job, and in media and broadcast, where show- and segment-specific vocabulary appears constantly.
Oak for Legal Sector
The customer-facing deployment that uses the workflow described in this article.
Frequently asked questions
What is a Jargon Library?
A list of your company's own vocabulary — product names, acronyms, internal jargon, client-specific terms, and people's names — that Oak uses when transcribing your meetings, so those terms come out right instead of being flattened into the nearest common word. You apply a library per project, and it's owned in your own workspace.
What should I put in a Jargon Library?
Five categories cover most of it: names (people, products, programmes), acronyms and their expansions, industry terms, internal jargon, and client-specific vocabulary. Anything specific to your organisation that a general model wouldn't have heard is a candidate.
How do I set one up quickly?
Seed roughly the obvious fifty entries first: your company name, top product names, the people who appear on your calls, and your most-used acronyms. A small number of recurring terms account for most mistranscriptions, so that first pass catches the majority of errors on day one.
How does the library stay up to date?
Add any term the moment a reviewer corrects it in a transcript. Because the same vocabulary recurs across meetings, the library becomes effectively self-curating within a quarter — it ends up covering your real working vocabulary without a dedicated effort to build it.
Will my confidential terms stay private?
Yes. A Jargon Library is owned in your own workspace, so a law firm's privileged terms, a media company's unreleased names, or a client's confidential vocabulary stay under your control and never become anyone else's. Applying a library per project also keeps a sensitive matter's vocabulary scoped to that matter. That ownership model is what makes the feature safe for legal and regulated teams.
Does a Jargon Library help with bilingual and Cantonese meetings?
Especially. Company-specific terms are exactly what a general model flattens in colloquial Cantonese, and the Jargon Library keeps them correct across languages. It's one of the foundations of reliable bilingual capture.