I was translating a large software project the other day and noticed at one point that I had mixed up the translations for words like file, folder and directory. Don’t ask me how that happened but by the time I noticed this the incorrect translations were all over the place and it would have been a time-consuming task to locate them individually since these terms were in almost every other segment. So I decided to utilize the QA checker to find the incorrect translations. This was easy to do with the Regular Expressions function, and the good news is that you don’t need to know or use any regular expressions to do this.
Go to Project Settings and select Verification > QA Checker > Regular Expressions. Select the Search regular expressions check box, if not already selected. Type a brief description or a name in the Description field. This is just for your own information. In this example we are trying to locate all segments where the word “file” is in the source but the Finnish translation does not include the matching term “tiedosto”, so as a description we can just use the word “File”. Type the source language word (“file”) in the RegEx source field and the target language word (“tiedosto”) in the RegEx target field. For the Condition, select Report if source matches but not the target from the pull-down menu. To save the search settings, click Action and select Add item. Create similar searches for other terms, as needed. That’s it and you can then close the dialog box by clicking OK.
Figure 1. Settings for a search for segments where the source text includes “file” but the target doesn’t include the matching translation “tiedosto”. Note the other similar searches for term pairs “database/tietokan” and “directory/hakemisto” below the “file/tiedosto” search.
When you run the Verification (F8), all segments where the source includes the term “file” but the target doesn’t include “tiedosto” will be flagged in the verification results. It worked beautifully in my case, and I had fixed the problems in less than 5 minutes. Another nice thing with this method is that it works well even with Finnish because you can just use the Finnish word stem without having to worry about the various endings the word might have in the text.
There were a few false positives caused by words like profile (the matching translation would be profiili). These were easy to skip while going through the verification results since there weren’t many of them. However, it’s also possible to fine-tune the search with the help of “real” regular expressions to look for exact matches only, if needed. You can also run the check in the opposite direction for extra security by using the Report if target matches but not the source option.