TMs, termbases & glossaries in the editor

Translation memories and glossaries attached to a project feed the editor while you work. A translation memory (TM) stores source–target pairs from past work so identical or similar text can be reused; a glossary pins how specific terms should be translated so terminology stays consistent. This page covers how those resources behave inside the editor. Creating and managing the resources themselves is covered under Translation memories & glossaries, and attaching them to a project under Translation setup.

TM matches

Pre-filled rows

When a project starts, rows whose source text already exists in an attached TM are pre-filled from the TM instead of being machine translated:

An exact match (100%) fills the row and sets its state to TM match.
A fuzzy match at or above the project’s threshold fills the row and sets its state to TM fuzzy.

The row’s state dot carries the match score in its tooltip, and the TM 100% / TM fuzzy row filters isolate these rows for review.

Applying a match while editing

For any row, the row state menu (click the state dot) lists the best matches from the attached TMs (up to two) as entries like Apply TM 100% — Client TM or Apply TM 87% — Client TM. Hover an entry to see the stored translation; click it to replace the target text with the match. Applying a match is a normal edit: it saves through the usual path and you can undo it.

The score is the similarity between your row’s source and the TM unit’s source — 100% means the source text is identical; lower scores mean you should expect to post-edit the differences.

Feeding the TM

The TM grows from your work: confirming a row writes its source–target pair into the project’s TM. Unconfirmed edits never enter the TM, so confirm only what you are happy to see reused.

The glossary panel

The glossary panel lives behind the round language button in the bottom-left corner of the editor. The badge on the button counts glossary terms matched in the current file; click it to open the panel. You can drag its corner to resize it.

The panel is organised into two sections:

In current segment — terms whose source form occurs in the row you are editing. This is the section to glance at while translating.
In file — every matched term in the file, with jump buttons.

Within a section, terms are grouped by glossary. Each term row shows the source term (with part of speech and domain when recorded) and the expected target term. The counters tell you how you are doing: the header shows how many matched terms already have their translation present in the target column, and each term shows its number of source occurrences — green when the expected translation is found in the target, amber when it is still missing.

Useful controls:

The paintbrush toggles term highlighting in the grid, so matched terms are painted in the source text as you scroll.
Click a term to select it and paint just that term’s occurrences.
The ▲ ▼ jump buttons (in the “In file” section) step through a term’s occurrences; the grid scrolls to each row and flashes the term.

Matching knows about word forms

Glossary matching is not a plain text search. In about 35 languages, matching is aware of word forms (lemmas), so an inflected occurrence still counts — a glossary entry for a base form matches its declined and conjugated forms in the text. Finnish gets particularly deep support, including compound words and consonant gradation. Glossaries marked case-sensitive are matched exactly.

Authoritative glossaries and forbidden terms

A glossary attached as authoritative (marked with a star in the panel) is enforced, not just suggested: a missing mandated term is flagged as a QA error on the row, and terms recorded as forbidden are flagged wherever they appear — with a one-click fix to the preferred term where the replacement is unambiguous. Non-authoritative glossaries surface the same information as warnings or reference only. Details in QA & entities.

The system termbase

Besides the glossaries you attach, the platform computes a system termbase for the language pair from public terminology collections (such as IATE and AGROVOC), matched to your file automatically. System termbase groups appear in the panel with a source badge and attribution. They are reference material — they are never enforced. If the panel says “Computing termbase matches…”, the match pass for the file is still running and the terms appear when it finishes.

Adding terms from the editor

You never have to leave the editor to grow a glossary:

Select the term in the source cell.
Select its translation in the target cell of the same row (the first selection stays highlighted).
Right-click the row and choose ＋ Add to glossary — the menu shows the pair you picked — or press Ctrl+Shift+Y (Cmd+Shift+Y on macOS).

The dialog opens with both fields pre-filled (you can also open it empty from the row state menu and type the terms). Pick the glossary, optionally record part of speech, domain and a comment, and add the term. New terms show up in the panel and in term highlighting immediately.

What you can do depends on the glossary:

Writable glossary — the term is added directly.
Read-only glossary — your term is sent as a suggestion to the glossary’s reviewers instead. Reviewers see pending suggestions at the top of the glossary panel and approve or reject them, optionally with a note.
No glossary attached yet — the dialog offers to create one for the language pair (with an optional case-sensitive matching mode) and attach it to the file in one step.

Glossaries — creating and managing glossaries
Translation memories — browsing and editing TM units
Translation setup — attaching resources to a project