Improving number + noun matches?

Question

Is there a way of tweaking the number + noun matches in Trados? One thing I struggle with a lot are sequences like 
 1 hr 
 2 hrs 
 3 hrs 
 4 hrs 
 5 hrs 
 10 hrs 
 11 hrs 
 12 hrs 
 Singular isn't a problem but once you move off singular, Trados appears to be very agnostic as to predicting what is a full match and what isn't, irrespective of the plural format for the target locale i.e. locales where the formula for plural is more complex than English (which has two forms; ONE and ELSE). Most languages have different plural structures - in my case, we have ONE (1, 11), TWO (2, 12), FEW (3-10, 13-20) and OTHER (anything else, including 0). 
 So I'd expect it, after 
 1 hr = 1 taigh 
 to show this as fuzzy 
 2 hrs = 2 thaigh 
 but then this also as fuzzy 
 3 hrs = 3 uairean 
 then these as full matches 
 3 hrs - 10 hrs = 3-10 uairean 
 then also as a full match 
 11 hrs = 11 uair 
 and so on. The problem is that when you get a list of these (even though I have tried and tried to turn off the way Trados automatically matches and moves forward one cell), before you know it you're at the end of the list and they're all wrong. 
 Is this a setting I can tweak somewhere or have plurals to date not made it into the gubbins of Trados? If not, it ought to be fairly easy to code this into Trados in relation to the target locale by simply pulling the plural formulae off CLDR I'd say.

Jerzy Czopik · Answer

If these examples are full segment content, I would't translate and autopropagate those. Just use display filter and enter ^\d+ hr$ (hr or whatever follows on your digits). This way you will have only segments like above shown. Search and replace hr/hrs and so on accordingly, select the segments and change their status to translated. This goes faster than you can imagine. In case this is a huge project, use SDLXLIFF Toolkit with same filter (this requires some knowledge in Regex in order not to destroy the SDLXLIFF, as there is no undo there). 
 
 BTW, I hope you indeed meant Studio, as you refer all the time to Trados - this is dead since 2009...

Paul · Answer

Hi Michael, 
 I think your choice of language was unfortunate since it demonstrates why it's not so simple to pull these from anywhere... I think? The link to the plural formulae doesn't have the example you quoted for example... I think is Scottish Gaelic but I am by no means sure. 
 Perhaps the best bet for now would be to use the regex autosuggest provider from the app store as this would allow you to construct the very table you wanted and have the correct plural ends suggested as you work. This wouldn't help with pretranslation of course but would help with interactive work. For pretranslation you could try the Terminjector which also allows you to create a set of rules and this might be better... depends how you want to work. 
 Studio 2017 might be able to solve this for you using the new Fragment Matching capability... I'll test this when we get the next Beta drop in due course. I'll let you know how well this works. 
 Regards 
 Paul

Paul · Answer

Unknown said: I'm not sure how I would do what you're suggesting, especially since the client expects there to be an 11 or 17 in the target string, not a formula. 
 Jerzy was referring to a regex search and replace.

Paul · Answer

Yep... so Regex AutoSuggest and Terminjector are your best options for now I think.

Paul · Answer

I have to admit, your usecase is quite tricky. For example, to be able to handle these: 
 
3 hrs - 10 hrs = 3-10 uairean 
 
You would need one rule like this: 
 
Find 
((?:10|[3-9]?[3-9])\s)hrs\s-\s((?:10|[3-9]?[3-9])\s)hrs 
 
Replace 
$1- $2uairean 
 
To handle just 3hrs or 10hrs and anything between you'd need something like this: 
 
Find 
((?:10|[3-9]?[3-9])\s)hrs 
 
Replace 
$1uairean 
 
Then for this: 
 
1 hr = 1 taigh 
 
Find 
\b(1 )hr\b 
 
Replace 
$1thaigh 
 
You get the idea perhaps... I think once you have all the rules, and maybe it's not too many, then the job would not be too difficult. But it certainly is a complex search and replace operations as the linguistic rules are not simplistic here. 
 
Regards 
 
Paul

Paul · Answer

Hi Michael, 
 
Like that it's less complicated. But you added even more complexity by asking for this: 
 
3 hrs - 10 hrs = 3-10 uairean 
 
As opposed to this: 
 
3 hrs - 10 hrs = 3 uairean -10 uairean 
 
A straightforward "number space hr(hrs)" is far less complex and I'm sure you could handle that with a few simple rules. 
 
Regards 
 
Paul

Paul · Answer

In that case it's pretty straightforward. The easiest way is probably as Jerzy said and allow Studio to do it's thing so you get all the numbers and hrs correct (not translated) and then do four search & replace operations. 
 Like this. 
 1 hr = 1 taigh 
 Search \b(1 )hr\b 
 Replace $1taigh 
 2 hrs = 2 thaigh 
 Search \b(2 )hrs\b 
 Replace $1thaigh 
 3 hrs = 3 uairean (in fact up to 10 hrs = 10 uairean) 
 Search \b((?:10|[3-9]?[3-9])\s)hrs\b 
 Replace $1uairean 
 11 hrs = 11 uair 
 Search \b(11 )hrs\b 
 Replace $1uair 
 So just four search and replace operations. Try it just by making sure you select the regular expression option when you search and replace: 
 
 I reckon when you've done it once or twice it won't seem so daunting any more. 
 If you wanted to do this interactively the same rules could be used in the Regex AutoSuggest Provider but personally I think for things like this the search and replace is better because you let Studio do it's thing quickly which will save you a lot of time, and then run a couple of search & replace operations across the file, or even an entire project at the end, in one go. 
 Regards 
 Paul

Trados Studio > 1. Trados Studio

Improving number + noun matches?