Under Community Review

Refine the measurement checks

Hi, Trados Community!

It would be of great help if the measurement checks (Verification > QA checker 3.0 > Numbers > Check measurements and currencies) was refined. Errors in units of measurements result in major/critical feedback from RI clients, with good reason - patients getting the wrong dose of the medicine, for instance.

The current in-built checker only checks measurements when they appear together with a digit, but that's not how all units of measurement appear in files: you can get, for example, a request to 'report the length of your rash in cm' or tables in which the table header includes the units of measurement so that each cell doesn't include it (this is standard practice in STEM), as well as axes, graphs, and a myriad other scenarios.

Furthermore, the current feature will warn that "measurement is missing in target segment or is not properly localised" where the source contains a number and a UoM but the target only contains a UoM: source: 5 cm/target: cm. In this case, this is a false positive as there is no problem with the units of measurement, only with the numbers. If both the numbers and units of measurement are activated in the Verify, then you get duplicated warnings. The number and unit of measurement checks should be independent, only getting a warning for numbers if there's a number mismatch; and only getting a warning for measurements if there's a UoM mismatch.

It also seems that the measurements aren't properly localised - is there a way to customise this? I have an en-es-ES file and it's showing a warning where the source is 5 g/L while the target is 5 g/l - in Spanish for Spain, the symbol for 'litres' is an upper case 'l,' so this is correct.

Diana Palenzuela

  •  

    I’ve voted up, but while improvements are implemented, you may want to go ahead and add customised regex rules.

    For example you can add the rule to check that if cm is in Source and not in target, it must be reported. The regex would be \bcm\b.

    Or you can combine as many symbols as you wish \b(mm|cm|ml|L|l|g|mmHg)\b or even full words (for example, if hour is in Source, there must be hora in the Spanish target; similar for seconds, minutes, days, weeks, months, years). 

    You can add rules for checking numbers too.

    About the issue with L and l symbols, I’m not sure about that, but I think that the measurement L expects L in target, while l symbol expects l. Again you can add a regex rule that if \b[Ll]\b is matched in Source, \bl\b must be matched in Target (case sensitive).

    I have it in that way, literally hundred of regex rules per target language, and some checks in Verifier turned off. It’s true that I get some unavoidable false positives, but really very few. A couple of manageable false positives is a minor collateral damage while assuring that days are not translated as years, for example.

    I hope it helps. More information in the following link: https://lalinternadeltraductor.org/n19/rule-based-qa.html

  • Sorry, small correction here as I don't seem to be able to edit: 

    It also seems that the measurements aren't properly localised - is there a way to customise this? I have an en-es-ES file and it's showing a warning where the source is 5 g/L (en-GB) while the target is 5 g/l (es-ES) - in Spanish for Spain, the symbol for 'litres' is a lower case 'l' and in English (UK) it's an upper case L, so this is correct.