DeepL

Justin Trumain wrote:

For example: DeepLcom translates "la Partie créancière de l'obligation" as "the Party owing the obligation." But in fact it means the exact opposite: the party to whom the obligation is owed (=the "obligee").

DeepL translates "Autonomie des stipulations" in a contract as "Autonomy of stipulations."
But this clause is called "Severability" in English.

[Edited at 2021-08-18 13:03 GMT]

Gerard Barry wrote:

Are you being serious? If all the translators of the world stopped working for a month, I can't see anything bad happening actually.

Tony Keily wrote:

"On certain more periphrastic legal or tricky financial/accounting documents, it's basically crap, but that's to be expected."

Juan David A Aquino Marquez wrote:

Hi,

Is there any software program, app, or web tool that can help differentiate between MT and Human-translated text?

Denis Fesik wrote:

I recently had an MTPE job that seriously pissed me off. It was a 50k-word construction standard, and I only had two days to make it seem like no MT had been involved (only a bunch of lousy human translators and maybe a good one or two). It all was a managerial SNAFU, so no big expectations there. What's remarkable, though, is that the MT, which had been trained using huge human-made TMs, didn't manage to translate any important terms correctly (and some of them had two or three incorrect translations: advanced AI sees a bunch of expressions where we see an interconnected body of text). That text was one big fail. Maybe there are language pairs in which a decent MT engine like DeepL can perform adequately, but it's certainly not the case with my working language pairs. Different languages often have hugely different systems for naming technical and scientific things, and English is one language where new terms keep popping up on top of huge piles of old ones, so whenever you have to communicate ideas from the frontiers of any discipline, MT won't help you – and dictionaries won't either. How will MT translate "simple and continuous construction"? Whatever it can come up with will be wrong. And none of the dictionaries I could find would offer correct translations. So, with jobs like that one, the only way to go is to open a zillion tabs with professional reference materials and get reading

Adieu wrote:

Still not happening. Unless you cut a lot of corners or it has a huge amount of exact reps.

Denis Fesik wrote:



It's just that someone had failed to organize things properly, and it wasn't me. They should give me time to complete the job with quality later. I covered a lot of ground building terminology for that project, but even after putting the correct terms together, I couldn't auto-find-and-replace them all throughout the text: it's a lot more time-consuming when editing Russian as compared to English. There was someone from (I guess) Italy on the forum who claimed he could do up to 15k per day starting with a clean slate, so maybe, given all the numbers and repetitions in that standard, he could have done an equivalent job. But not me, there's no way I can get that fast

Adieu wrote:

MTPE translator role? Or revision role after MTPE was performed by someone else?

If it's revision... it's remotely possible with dogged determination and a case of energy drinks.

Gerard Barry wrote:

I'm currently post-editing a contract that was translated by DeepL from German to English. I find the quality really good!

Denis Fesik wrote:



Legal English is full of things like "To improve, manage, control, cultivate, develop, exploit, exchange, let on lease or otherwise, mortgage, charge, sell, dispose of, grant as gift, turn to account, grant rights or privileges in respect of, or otherwise deal with all or any part of the property, assets and rights..." – and I'll bet all my money (and will even borrow to bet some more) that no AI, however advanced, will translate this into a grammatically correct sentense in my language, because over here, every verb in the list calls for a specific noun case governed by it. AI will just translate the verb list and decline the nouns according to whichever verb comes last. The problem is that most human translators do the same thing, but I'm not one of them. I wonder how DeepL handles German cases in similar situations (four cases with hardly any noun forms are an easier task), but my overall observation is that German works in ways very different from those of English, which is why a lot of texts written by Germans in English are sometimes completely incomprehensible, but when you get a chance to read the original German, you begin to understand what the English was supposed to mean. So far, I've only been talking about linguistic things, but communicating civil law things to someone with a common law mind frame is yet another problem AI cannot solve, it does call for some ingenuity

Gerard Barry wrote:

I'm currently post-editing a contract that was translated by DeepL from German to English. I find the quality really good! The only issue I have is that my dumb department (I'm an in-house translator) insists not only on post-editing the DeepL translation but also having it proofread by another translator after that. To me, this seems really excessive. What's the point of using technology if you have to check everything twice afterwards anyway?