Translation memory
A translation memory is a database of previously translated segments, paired source and target, that is reused on new work. It keeps terminology and phrasing consistent across projects and reduces cost, because repeated or similar content does not have to be translated again. It is a valuable asset, and who owns it matters.
How it works
As a translator works, each approved segment, a sentence or phrase paired with its translation, is stored in the memory. On the next job, the tool checks new text against it: exact repeats come back as full matches, close ones as fuzzy matches with a percentage, and only genuinely new content has to be translated from scratch.
The memory grows with every project, so it gets more useful over time. It also enforces consistency, because the same source sentence returns the same approved translation rather than a fresh guess.
How SourceTarget uses it
SourceTarget uses translation memory to keep terminology and phrasing consistent across a client's work and to avoid paying to translate the same content twice. Matches are surfaced during the job and reused where they hold up.
Ownership is explicit: the memory built from a client's content belongs to that client, so the value accumulated stays with them.
Translation memory compared with Termbase
| Translation memory | Termbase | |
|---|---|---|
| Stores | Whole translated segments | Approved terms and their translations |
| Unit | Sentences and phrases | Words and short phrases |
| Purpose | Reuse and consistency of text | Consistency of terminology |
| Grows by | Every segment translated | Curated term entries |