How to Use the Translation Glossary

A well-maintained glossary is the most effective way to ensure AI translations use your exact terminology every time. You can build it manually, import an existing term base, or let the AI generate entries automatically from your localization files — all three methods work together.

1. Create a glossary: Click New Glossary, give it an optional name, then select the source language (e.g., en-US) and target language (e.g., de-DE). Enable Active to start applying it immediately.
2. Add entries manually: Open the glossary and click Add Entry. Enter the source term, its approved translation, and an optional context note to clarify usage — for example, "bank" as a financial institution rather than a riverbank.
3. Import in bulk: For large term bases, click Import CSV/TSV and upload a file with columns: source, target, and optionally context. This lets you migrate existing glossaries in seconds.
4. Generate entries with AI: Paste your source strings — or both source and already-translated strings — into the Generate Glossary with AI section and click Generate Glossary or Generate Glossary Entries. The AI analyses your content, extracts recurring terms, product names, and domain-specific vocabulary, and populates the glossary automatically. Existing entries are taken into account to avoid duplicates.
5. Translate with term enforcement: Any active glossary whose language pair matches your translation is applied automatically — no extra steps. Run your translation as usual and the AI will respect your approved terms.
6. Export for backup or sharing: Click Export CSV or Export TSV to download your glossary. Use it as a backup, share it with colleagues, or import it into another workspace.

What Is a Translation Glossary?

A translation glossary is a controlled vocabulary that maps specific source-language words or phrases to their approved target-language equivalents. Unlike a general dictionary, a glossary reflects deliberate editorial decisions for a particular domain, brand, or audience — ensuring the AI model uses the exact terminology you have chosen rather than a valid but unintended synonym.

For example, a software product might require that the term "Settings" is always translated as "Einstellungen" in German rather than "Konfiguration" or "Optionen". A legal firm might enforce precise equivalents for contract terms to prevent ambiguity. A healthcare provider might lock in clinical terminology where accuracy is non-negotiable.

Glossaries are especially valuable when working with AI translation, where the model may otherwise choose from several equally valid synonyms. Providing explicit term constraints guides the model toward the vocabulary your users and stakeholders expect.

How Glossary Enforcement Works

When you run a translation, l10n.dev automatically identifies all active glossaries whose language pair matches the current translation direction. Your term pairs are supplied to the AI as constraints — the model is instructed to use the specified target term whenever the corresponding source term appears in the text.

Terms are matched case-insensitively and applied in context. Use lowercase for general terms; the AI will adapt capitalisation to the sentence context automatically. Use exact capitalisation only for brand names, acronyms, or case-sensitive terms. If the same surface form appears in multiple senses, adding a context note to the glossary entry helps the AI apply the mapping only when the intended meaning is present.

Context Notes

Each glossary entry supports an optional context note — a short description that disambiguates the term. Context notes are especially useful for polysemous words (words with multiple meanings) and for distinguishing domain-specific usage. For example, adding "financial institution" as the context for "bank" prevents the glossary from overriding that word in unrelated sentences.

CSV / TSV Import and Export

Glossaries can be managed in bulk. The import format expects a header row with at least source and target columns; a third column named context is optional. Both comma-separated (CSV) and tab-separated (TSV) formats are accepted. Use the Export CSV or Export TSV button on any existing glossary to download a ready-to-edit file.

Common Use Cases

Translation glossaries are valuable wherever precise, consistent terminology matters across languages:

• Software UI localization: Lock in the translations for UI verbs and nouns — e.g., always render "Submit" as "Soumettre" in French rather than "Envoyer".
• Legal documents: Ensure contract terms such as "indemnification", "liability", or "covenant" are always rendered with their approved legal equivalents.
• Medical and clinical content: Maintain precise clinical terminology — for instance, distinguishing "myocardial infarction" from "heart attack" in contexts where specificity is critical.
• Brand voice and product names: Keep product names, feature labels, and brand-specific terms consistent across every language and market.
• E-commerce catalogs: Standardize product attributes, category names, and unit labels across large item catalogs translated into multiple languages.
• Marketing and campaigns: Preserve the approved translations for slogans, calls to action, and campaign names so messaging remains coherent across channels.
• Technical documentation: Lock in the agreed translations for API terms, error messages, CLI commands, and field names so developers and end-users encounter the same vocabulary everywhere.

Generate Glossary Entries with AI

L10n.dev can automatically build your glossary from existing localization files. Paste your source strings — or both source and already-translated strings — and the AI extracts recurring terms, product names, and domain-specific vocabulary, saving them as glossary entries ready to enforce consistency in future translations.

AI glossary generation works with any supported localization format: JSON, YAML, PO, ARB, XLIFF, and more. The model considers already-existing entries to avoid duplicates and conflicting terms, so you can run generation multiple times as your content grows.

Frequently Asked Questions

How many glossaries can I create?

You can create as many glossaries as you need — for example, one per language pair, one per project, or one per product line. All active glossaries whose language pair matches a translation are applied at the same time.

What happens when multiple glossaries match the same translation?

All active, matching glossaries are applied together. If the same source term appears in more than one glossary with different target translations, the AI uses context to resolve the conflict. To avoid ambiguity, keep one canonical entry per source term per language pair.

Are context notes required?

No. Context notes are entirely optional. For unambiguous technical terms a plain source → target pair is sufficient. For polysemous words — words with multiple meanings — a brief context description (e.g., "bank — financial institution") significantly improves translation accuracy.

What is the required CSV import format?

Expected columns (header row is optional): source, target. A third column named context is optional. Both comma-separated (.csv) and tab-separated (.tsv / .txt) files are supported. Click Export CSV or Export TSV on any existing glossary to download a correctly formatted file.

Can I temporarily disable a glossary without deleting it?

Yes. Every glossary has an Active toggle. Inactive glossaries are preserved in your account but are not applied during translation. You can re-enable them at any time.

Does a glossary apply to all language pairs?

No. Each glossary is scoped to a single source language and a single target language. You can create separate glossaries for different language pairs — for example, one for English → French and another for English → German — each with its own tailored term list.

What does the character count shown in a glossary represent?

The character count is the total number of characters across all source terms, target terms, and context notes in the glossary. Because glossary content is included in the AI prompt during translation, this figure indicates how much the glossary contributes to your overall character usage per translation request.