Reverse a TMX or extract one language pair
A German → English project needs your English → German TM. A client sends one TMX with 24 languages and your TM in Trados takes one pair. Both jobs are the same file with a different idea of what the source is.
━━ the problem
Once German is the source, two English sentences that were translated to the same German become duplicates — and the header, each unit's srclang and the order of variants all have to follow.
Trados and memoQ import one language pair per TM. Feeding them a 24-language export imports one pair at best and leaves you wondering where the rest went.
Extracting "English" from a multilingual TM should take every English variant, and extracting en-GB should take just that one. Most tools only do one of the two.
━━ how TM Cleaner handles it
Drop the TMX and open the “Will it import?” check under the result. For a two-language TM, “Reverse” makes the other language the source; for a multilingual one, tick the target languages to keep. Update the preview: duplicates and every rule are judged again on the new direction, and the cleaned TMX gets the new header and order.
- ✓reversing re-judges duplicates on the new source, not just the column order
- ✓“fr” keeps fr-FR and fr-CA; “fr-CA” keeps only fr-CA
- ✓units without the chosen languages are listed under “other languages”, not silently lost
- ✓removed.tmx keeps the original direction, as an undo
━━ frequently asked
can I reverse and extract at the same time?+
yes: choose German as the source and keep only French, and a 4-language TM becomes a German → French one.
what happens to units that don't have the language?+
they're removed, listed with their reason, and kept in removed.tmx.
does the TSV / Excel export follow the new direction?+
yes: column one is the new source.
━━ related pages
free · no account up to 50 MB · files up to 2 GB · try the demo