The translation memory (TM) is a collection of approved translations. When a new text looks like one that was translated before, the earlier translation is offered as a starting point, so the same sentence is not translated twice.
What goes in #
An entry is created or updated every time a translation is approved: by a reviewer, by publishing when the project does not require review, or by an import that does not mark translations for review. An entry stores the source text, the translation, the language, the project, the key and the branch, who approved it, and how often it has been used.
Approving the same key and language again replaces the earlier entry. When a branch is merged, its entries move to the base branch.
Where it is used #
In the editor #
For the key you are working on, the editor shows up to five TM matches of at least 50% in the Translation Memory panel. Each match shows:
- the match percentage,
- the source text and key it came from,
- its translation,
- the language, how often it was used and the project.
Choose Apply to put the translation in the field, or press M to apply the best match. Using a match raises its “used” counter. Matches come from all projects you have access to, but not from the key you are editing. Reviewing and saving stay up to you. See The translation editor.
On the Translation memory page #
Open Translation memory in the project navigation to browse and search the whole memory.
static/screenshots/translation-memory-page.png- Type a text in Search translation memory and choose Search. Results are ranked by how similar their source text is to your text.
- Without a search, the most used entries are listed.
- Pick the language, and limit the results to one project or All projects.
- The figures at the top show the total number of entries, language pairs, times reused and projects.
- Select an entry to see its details: project, who approved it, when, how often it was used.
- Copy puts the translation on your clipboard and counts it as a use.
- Delete removes an entry from the memory. This needs the reviewer role.
- Export TMX downloads the listed entries as a
.tmxfile for other translation tools.
Match levels #
| Match | Label |
|---|---|
| 95% and higher | Exact |
| 80% to 94% | High |
| 50% to 79% | Fuzzy |
| 30% to 49% | Low |
| Below 30% | Poor |
The percentage is the similarity between the source text you search for and the source text of the entry.
Access #
You only get entries from projects you have access to. Deleting an entry does not change any translation, it only removes the suggestion.