Glossaries

A glossary pins how specific terms should be translated, keeping terminology consistent across a project or client. Each entry is a term pair — a source term and its preferred translation — optionally enriched with usage metadata. Unlike a translation memory, which reuses whole segments, a glossary works at the level of individual words and phrases.

A glossary has a language pair but applies in both directions by default: an English ↔ Finnish glossary serves English → Finnish and Finnish → English work alike.

Creating a glossary

Import a TBX, CSV, or tab-delimited TXT file on the Resources page. Languages are detected automatically and can be corrected afterwards.
Create an empty glossary — on the Resources page or during translation setup — and add terms as you work.

Each glossary has a case sensitivity setting (off by default). Leave it off for ordinary terminology; turn it on when casing separates your term from a common word — “Jaguar” the brand versus “jaguar” the animal.

The glossary viewer

Click a glossary card to open the viewer: a table of term pairs, loading incrementally, with the same search toolbar as the TM viewer (scope the search to source, target, or both; filter to matches).

Edit a term by double-clicking either cell; Enter saves, Escape cancels.
Open the detail panel to edit a term’s metadata:
- Part of speech — noun, verb, adjective, and so on.
- Domain — the subject area the term belongs to.
- Forbidden term — marks a rendering that must not be used (see below).
- Comment — free-form usage notes for translators.
Remove a term from the right-click menu (or the row’s menu button).
Download TBX exports the whole glossary.

How terms surface in the editor

When a glossary is attached to a project’s language pair, the editor loads exactly the terms that actually occur in the open file — a glossary can hold a million terms and the editor still only handles the handful that matter. Occurrences are highlighted in the source text, and the terms for the current segment are listed in the terminology panel. Matching is whole-word, and it respects the glossary’s case-sensitivity setting.

Read-only reference termbases can appear in the same panel alongside your glossary terms; see TMs, termbases & glossaries in the editor.

Inflection-aware matching

For about 35 languages, term matching is lemma-aware: a term is recognized even when it only appears in an inflected form. Finnish gets the deepest coverage, with full morphological analysis that handles compound words and consonant gradation — so the term “sopimus” is found and highlighted in a sentence that only contains “sopimuksen”, and clicking the inflected word takes you to the term. If analysis is unavailable for a language, matching falls back to plain text matching.

Adding terms from the editor

Translators can add term pairs without leaving their work. The add-term dialog in the editor lets any project member add to their own glossaries, or to a glossary the project manager has shared with the project (as long as it is not marked authoritative). Terms added in project context join the glossary immediately but stay withheld from exports until the project is fully paid — see withheld additions.

Terminology checking

Glossary terms feed the automated quality checks that run on every saved translation (see QA & entities):

Missing term. If a source segment contains a glossary term but the translation does not contain its preferred rendering, the segment is flagged — as an error when the glossary is authoritative, otherwise as a warning. When a glossary lists several accepted translations for the same source term, any one of them satisfies the check. Inflected renderings count as present in languages with inflection-aware matching.
Forbidden term. A term marked forbidden inverts the rule: that rendering must not appear in the translation. Using it — in any inflected form — is always flagged as an error. Each violated term gets its own finding, so dismissing one never hides another.

Finnish inflection-aware autofix

For Finnish targets only, a forbidden-term finding can carry a one-click fix. When the forbidden word appears in an inflected form and the glossary has exactly one preferred alternative, the fix replaces it with the preferred term inflected into the same grammatical case and number — not just the dictionary form. If the system cannot generate the inflected form with confidence, the finding is still shown, just without the automatic fix — it never guesses grammar.

Translation memories & glossaries — importing, exporting, metadata, and attaching.
TMs, termbases & glossaries in the editor — the terminology panel and highlighting.
QA & entities — how findings are shown and resolved.