Daily Mail escapee?
Good Lord... if I want to read that kind of unbalanced tosh, I'll go to the Daily Mail website.
The European Commission is seeking to make us all speak in Brusselsese by donating millions of its documents to translation software developers. The commission described the donation of its "collection of about one million sentences and their high quality translations in 22 of the 23 official EU languages" as a step further in …
Stop Press: EU WRITES LEGAL DOCUMENTS IN LEGAL LANGUAGE!
Register Journos Shocked:
Textual Detail Of International Treaty Is Difficult To Understand!
Give 'em a break. Compared to most international legal stuff, the EU treaties are a model of lucidity. Try this, from the Kyoto Protocol:
"If a question of implementation by a Party included in Annex I of the requirements referred to in this Article is identified in accordance with the relevant provisions of Article 8, transfers and acquisitions of emission reduction units may continue to be made after the question has been identified, provided that any such units may not be used by a Party to meet its commitments under Article 3 until any issue of compliance is resolved."
Besides, it's member countries like our own dear Britain (not "Brussels", whatever that means) who write the European treaties. And when you've got 27 governments wrangling for weeks over the precise positions of full stops and commas, legal "gibberish" is the guaranteed outcome. Better that than vague arm-waving and leaky prose.
(And "Brusselsese"? Please...)
You're making the classic mistake of double-counting language pairs. You count 22 pairs for, say, English, then 21 for French (because you've already counted French-English), then 20 for German (because you've already counted German-English and German-French), etc etc. Two seconds in a spreadsheet or several more with a calculator gives you 253 pairings between 23 languages.
"So half of the EU isn't talking to each other?" - If only, can we all stop speaking to the French?
I think the maths is vindicated by the word "combos" meaning combinations, so English-German and German-English are treated as one combination, rather than two.
23 x (23 - 1) / 2 = 253
-1 since we don't translate a language into itself
/2 to give the number of language pairs (ie English -> French, French -> English counts as 1)
I once met a very nice lady, who was doing simultaneous English <-> French for us. She is Danish married to a Portuguese, and is therefore one of the (very) few professional translators available for those two languages. I believe some of the more obscure language pairs have to go through an intermediate language, which can't improve the accuracy of the final result!
Irish just recently became promoted as an official EU language, so they are still lagging behind with the translations of all the important documents.
Welsh, Catalan and other minority languages are not official EU languages, which is a shame, as translating EU documents significantly improves translation and language studies in all the small language groups. I come from Slovenia, which has a language with only 2 million speakers and in the past few years there has been a real explosion of translation jobs and a noticeable increase in the quality of work done.
English (and sometimes French) are used as bridge languages by the EU since 2004, but only for simultaneous interpreting (in parliament, conferences, etc) - there aren't enough simultaneous translators trained in those languages. I don't think bridge languages are generally used for written documents, and certainly not for legal stuff.
It is not europhobia to question the value of european documents.
The Kyoto protocol cited above reads to me like legalese. Joe's examples read like translations.
EU translations have a widely acknowledged problem: each translation must be utterly unambiguously identical to the original. This is essentially impossible. Translators will sometimes invent new words because they can't change the sentence structure for fear of changing meaning. Reading an EU document is not really reading "in your language", but it does generally provide the intended information very effectively.
However, Joe misses one vital point.
Anyone involved in translation knows this. Even undergrad language students know this. Which means the people that write this software will be more than aware of this.
Panic over -- move along -- nothing to see here.
Google are busy building a database of translation pairs using user submitted translation suggestions. If they're going to submit that corpus of texts to Google, they should also get a quid pro quo agreement from them.
We have major companies working in that field in the EU (e.g. Systrans) to release that body of text without a balancing agreement would tip the field in Googles favour.
i.e. we paid for it, make sure you get a balancing agreement before you go giving it away.
At least we don't have too much of a problem here. Unfortunately, other languages creep into our country (it happens!). In many places we have "Spanish Subtitles" to stuff (I note it in banners in a couple of home improvement stores). Luckly there are only a few dominant languages: English (more rightly "american"), Spanish, and (luckly only ONE country has it as a sole language!) French. I suppose we are lucky.
Unfortunately, the French won't give up on their language being "pure", which limits things a bit. There will always be "The French"!
It is a rule in translation that you only translate *into* your own language. (you only have to read foreign translated instruction manuals to see why this is a good idea.), so for each language pair, you really need two translators - one for each native language. However, there is no shortage of translators in modern Europe, so finding them - at least for the major languages - is not a problem
While politically, it might be necessary to have EU documents translated into Gaelic, practically, most Irish understand English as well as us natives (if not better,) certainly far more than understand Gaelic
James Joyce, Oscar Wilde, George Bernard Shaw, Spike Milligan, et al, bear witness to that facility with our mother tongue.
Translation software that's any good doesn't provide an out-of-the-spout final version. It provides useful suggestions for words, phrases and sentences, and the translator selects what's appropriate and puts it all into acceptable shape. Translations into English should sound English, etc. Nothing tires or puts a reader off more than non-idiomatic or inconsistent usage. This even goes for stuff that's crap in the original - the translation should read like English crap etc unless you're being paid extra to de-crap it.
That's why this kind of offer is a great step forward. It provides real translators with real suggestions provably used in real cases. Nothing there to stop anyone coming up with anything better.
Out-of-the-spout versions are very hit and miss. The only people who can really make full use of them are those who know the original language and can decipher some of the oddities arising from over-literal translations and misunderstood contexts. More than a page of this for a stranger to the original language is usually mind-numbing enough to bring the reading to a halt, unless you're driven by an extraordinary passion for the topic.
There are a couple of problems with huge corpuses like this, though, that have nothing to do with gobbledegook. The first is the need for constant cultivation - pruning, editing etc, to weed out possible errors and focus on the useful alternatives so that a translator isn't overwhelmed by a deluge of not very useful suggestions. Another is the need to keep terminology precise and up-to-date. A third is more technical - a capability to segment a sentence into useful bits and pieces so that phrases and words are matched rather than just full sentences. Translators who used DejaVu as against Trados translation memory software are constantly thankful for the greater flexibility DV offers in this respect.
Finally, we should be aware of the de-skilling of translation as a profession that is going on. Trados software in particular is time and again used by translation agencies to gouge translators and push rates down. On the other hand, there are still one or two clients around who appreciate that a good translation is worth its weight in gold. They don't want monkey jabber, so they don't pay peanuts. (Note for the business-minded - the time taken to make a crap translation acceptable usually exceeds the time needed to get a good translation in the first place, if the translators know their onions.
"Stop Press: EU WRITES LEGAL DOCUMENTS IN LEGAL LANGUAGE!"
Well if they spent some time simplifying the original (*) text, they'd save a bundle on translation costs and saves years of subsequent wrangling over treaties that meant something slightly different to every country.
IT angle: it's a classic case of automating a flawed process.
Well in my case (I'm the one that defended the author... to a point) it's because I foolishly registered with my own name, thinking that was the done thing here.
Now I see the error of my ways, as these pages are filled with comments from pseudonymous accounts like "Hate2Register".
I can't see any account settings, so... aw heck -- why not go AC?
Occitan in the South East would be a bit of a turn up for the books, I would expect to find it a long way West of there - West of the places where Provencal is still alive (or was last time I was there - a few decades ago) albeit not very alive.
It would be nice to have translations of EU documents into all those minority languages from France, and or course into Gaoluinn and Gaeilge and Gaidhlig and Gaelk and Welsh, also Cornish if there's even one person who speaks it well enough to translate this sort of stuff. Maybe I could get a job as a translator and give up this IT nonsense.
Biting the hand that feeds IT © 1998–2021