I'm currently retranslating a text that a non-translator colleague had partially translated, and I noticed an abundance of double genitives in both source text (DE) and target text (EN) (rendering of German "des/der ... des/der ..." formulations with "of ... of ...") and in wanting to understand more about Regex was wondering if there might be a quick and dirty Regex-based QA check that could be done to trap these kinds of double genitives.
Does anyone have any suggestions on this?
Can you provide a couple of sample source and target sentences you need to catch?
Paul Filkin | RWS
Design your own training!
You've done the courses and still need to go a little further, or still not clear?
Tell us what you need in our Community Solutions Hub
It seems to be particularly prominent in titles and positions (e.g. in tabular reports of names and positions)
Example 1:
Example 2:
Similarly I have a very lengthy list of titles, where some have been been translated as A of B of C (C is usually a company name)
Sometimes it starts getting silly.
At a very simplistic level perhaps something like this would suffice?
\bof\b.*(?=(?:\bof\b))
You could use this as a QA check in the target for example, or even in the display filter to filter all the segments that do this.
Paul Filkin | RWS
Design your own training!
You've done the courses and still need to go a little further, or still not clear?
Tell us what you need in our Community Solutions Hub
Your question is interesting because it is also one of the examples in my book.
You might want to refine the regex to match only occurrences of "of the" (and "of") near each other, such as
\bof the(\s*\w+){0,2}\s+of the\b
that finds "of the" separated by less than 3 "words". The bounds could be adapted to be less restrictive, but as Paul has often said consider "economy of accuracy" - how much effort do you want spend on refining your regex and how many false positives/negatives are you prepared to accept.
Your question is interesting because it is also one of the examples in my book.
You might want to refine the regex to match only occurrences of "of the" (and "of") near each other, such as
\bof the(\s*\w+){0,2}\s+of the\b
that finds "of the" separated by less than 3 "words". The bounds could be adapted to be less restrictive, but as Paul has often said consider "economy of accuracy" - how much effort do you want spend on refining your regex and how many false positives/negatives are you prepared to accept.
Thank you for the addition to the regex to help with proximity - I tried a few values from less than 3 words to 6 words, which caught everything that I wanted to catch - and meant that I was able to get through the list!
Why the asterisk after the first space?
Shouldn’t it be
\bof the(\s\w+){0,2}\s+of the\b
instead?
True, it was a typo - it should have been
\bof the(\s+\w+){0,2}\s+of the\b
But as I mentioned, the bounds may need to be customised
Translate
