🤬 When will RWS (formerly SDL) Trados and Wordpress 's WPML finally get their sh*t together and make translation of WPML's #XLIFF dialect less of a PITA instead of keeping to point towards the respective other, like they've done for years?
Now, for reasons unknown, Trados' "Multilingual XML" file filter, which was supposed to improve things, escapes the entire CDATA'ed HTML upon export. 🤦♂️
I'm almost considering #OmegaT once again…
TIL "they" as a singular pronoun is *way* older than I thought: ~ 1375 AD.
Source: https://www.scu.edu/media/offices/provost/writing-center/resources/Tips-Singular-Pronoun-They.pdf
🤔 Now, is the fact that next to "statuses", "status" is a correct plural of "status" in #english ...
Liebe #InDesign #DTP -Menschen, bitte schickt eurem Übersetzer kein PDF. Unsere CAT-Tools verwandeln das für die #xl8 in ein meist layout-zerschossenes DOCX, das ihr per Copy/Paste nach ID zurücktragen dürft.
Exportiert stattdessen ein .IDML (ID's XML-Austauschformat), das ihr nach der Übersetzung nahtlos nach ID zurückimportieren könnt, dann wird die Welt für uns alle einfacher. Danke. #servicetoot
🇬🇧 to 🇩🇪 #xl8, where the source text is all too obviously a 🇫🇷 to 🇬🇧 translation.
Working also with 🇫🇷, I happen to know that #French puts adjectives in plural if they refer to a series of 2+ nouns, *even if* they are all in singular.
But "Critical risk control measures … and prevention" is *not* the same as "Critical risk control and prevention measures".
TL;DR: 15 years into the job, chain translations are still a bad idea (as is leaving machine translation unsupervised).
I am sure there exists a special place in hell for people translating 🇫🇷 "alcool, drogues, médicaments" as 🇬🇧 "alcohol, drugs, drugs" #xl8
https://www.spiegel.de/start/dolmetscherin-werden-oleksandra-popek-erzaehlt-von-ihrem-ersten-jahr-im-job-a-4a77f64f-4b68-4096-aa85-db6552e0c776 #1nt
The @Wired magazine has made a well-rounded overview of what an interpretation AI can and cannot do at this point:
https://www.youtube.com/watch?v=pwOxlpGYJAY
Most of this applies to machine translation, too, because in fact, #1nt #ai is a speech-to-text -> #mt #xl8 -> text-to-speech toolchain.
They are tools, good for some use cases, bad for others.