Translation Memory

Translation memory (TM) stores every confirmed translation as a source-target pair. When you translate a new segment, Sygaea searches the TM and suggests previous translations that match. This helps you maintain consistency across a project and speeds up work on repetitive texts.

How to use translation memory

Viewing TM matches

1. Click on a segment in the editor to make it active.
2. Sygaea automatically searches the translation memory for the active segment’s source text.
3. If matches are found, they appear in the Matches tab of the bottom panel.
4. Each match shows the source text, the stored translation, and a match percentage indicating how closely it matches.

Inserting a TM match

1. Review the matches in the bottom panel.
2. Click on a match to insert its translation into the target field of the active segment.
3. The inserted text replaces any existing content in the target field.
4. Edit the inserted text as needed to fit the current context.

Inserting a TM match is a starting point, not a final translation. Always review the inserted text to ensure it is appropriate for the segment you are translating.

Using concordance search

Concordance search lets you search the entire translation memory for a specific word or phrase, rather than waiting for automatic matches.

1. Switch to the Concordance tab in the bottom panel.
2. Type a word or phrase into the search field.
3. Press Enter to search.
4. Results show TM entries where the search term appears, with the matching text highlighted in context.
5. Click on a result to insert its target text into the active segment.

Concordance search is useful when the automatic TM matches do not find what you need, or when you want to see how a particular term has been translated elsewhere in the project.

Translation memory reference

Match percentages

TM matches are scored by how closely the stored source text matches the active segment:

• 100% match: The source text is identical. The stored translation can likely be used directly with minimal editing.
• Fuzzy match: The source text is similar but not identical. The percentage indicates the degree of similarity. Lower percentages require more careful review and editing.

Match scoring uses a multi-pass approach that accounts for orthographic normalisation (handling spelling variations in historical texts) and token-level comparison.

How TM entries are created

TM entries are created automatically when you confirm a segment. Each time you press Enter to confirm a translation, the source text and target text are stored as a TM entry. This means the translation memory grows organically as you and your team work through documents.

Multiple translation memories can be linked to a single project. This allows you to share TM resources across related projects or use a curated TM alongside one that accumulates new translations.

Auto-fill for exact matches

Projects can be configured to automatically fill target segments when a 100% TM match is found. When enabled, opening a document will pre-populate any segments that have exact matches in the linked translation memories. The project lead can enable or disable this feature in the project settings.