Best practices for reviewing UI and help translations using Apsic Xbench 3.0

Here are my best practices for reviewing UI and help translations in software localization using Apsic Xbench 3.0, Duden Korrektor Plus and text-to-speech software and for implementing client feedback to constantly improve translations.

I have been working as a freelance translator and reviewer in the localization business for about 9 years. Translated words up to now: around 2 million words, reviewed words up to now: around 900.000 words. With this background, I was able to gather some routine in this area. My reviewing routine is something I want to share with interested people working in the same area. Maybe this is of help for someone who is struggling to streamline his or her reviewing workflow. Here is mine.

Step 1: Xbench QA
  1. Create a new project in Xbench. Add all the files you want to review.
  2. QA files in Xbench.
    Optional: I usually do not do a check for keyterms mismatches as this usually yields a lot of false positives. If time permits though and the quality of the translation indicates there might be issues here, I do a check for keyterm mismatches as well.
  3. Implement changes for all indisputable errors. I collect my changes in a Word file: Copy implemented segments to a word file (to be sent to translator for acceptance). Also copy segments that need translator feedback to this Word file.

I am using a Word template for this. It contains a checklist and four list items: Nice solution, Implemented, To check, Translator feedback necessary.

After having implemented changes for a certain segment, mark it done in Xbench by pressing Ctrl+M.

Step 2: Duden Korrektor Plus Spell Check in MS Word

Export the translated files to a Word file: Click Ctrl+W in Xbench to export the file (exclude ICE matches and duplicate sentences if needed). Press F7 in Word to start spell check. Implement changes in the translated file. Copy erroneous segments to review template file as a reference for the translator.

Step 3: Listening spot check with TTS software Linguatec VoiceReader or Chrome TTS plug-in

Using this approach, I often find mistakes that would otherwise have slipped under my radar. Listening to text helps to find grammar issues that passed the spell check.

Highlight the entire target text or parts of it. Export it to an MP3 file, listen to it with F4, a transcription software that allows easy forwarding and rewinding of audio files (which unfortunately VoiceReader is not capable of).

Step 4: Send change log to translator

If time permits, I send a changes log to translator and ask him to approve changes. I wait for translator feedback, implement any necessary changes and deliver the file to the client.

Xbench 3.0 advantages

  • You can easily remove ICE matches and duplicate strings for proofreading (in case you are not using a sophisticated tool like SDL Studio).
  • You can export all files in a project to ONE single Word file for proofreading.
  • Jump directly to a sentence/string in a specific file for quick change implementation (Ctrl+E). This is especially useful if you need to implement changes across a large number of files.

My best practices

  • Implement major errors from LQE reports and updates from client as Xbench rule to avoid them in the future.
  • Create own rules in Xbench to reflect client feedback.
  • Use word macros to spot errors directly in MS Word, e.g. highlight blacklisted words like Anweisungen.
  • Use LQE feedback from the past as quick reference in XBench (save them as tab separated txt file).


Add German corpora from to your Xbench project for quick German usage reference (shortcut: Ctrl+Alt+Backspace)

Hands-on task: Simple RegEx QA rules in Xbench

  • Create a QA rule in Xbench to automatically spot a blacklisted word.
  • Create a QA rule in Xbench that checks if there is a non-breaking space between z. and B.
  • Create a QA rule in Xbench that checks if mobile device has always been translated with Mobilgerät.