I know that Deepl is now used for sentence translations but it doesn’t work well with Asian languages such as Korean.
I just came across this sentence tonight:
서너걸음 거리를 두고 걷는 그들의 틈새로 꺼져버린 불의 잔해에서 피어오르듯 안개가 엷게 몰려왔다.
Deepl: As they walked three or four paces apart, a thin mist rose from the remains of the extinguished fire.
Google Translate: As they walked three or four paces apart, a thin fog rolled in between them, as if rising from the remains of an extinguished fire.
Deepl gets it wrong here and is really struggling with Korean complex sentences. Even if I believe language learners shouldn’t resort to translating sentences systematically, inaccurate translations can act like a spanner in the wheels.
Was curious to see what a Korean website would do.
Papago Naver: The fog came thinly as it rose from the wreckage of the extinguished fire through their crevices, walking three or four steps away.
Ugh. Even worse than Deepl. People who work as Korean to English translators will have their jobs for a very long time.
Thanks for your suggestion, I’ll forward this to our team.
It would be awesome to have a choice but I guess it’s difficult to manage. I would chose Google Translate over DeepL as well.
GPT to rule them all?
As they walked a few steps apart, a thin mist seemed to rise from the ashes of the extinguished fire, creeping in between them.
Except it’s wrong. There is no actual fire here. Just a mist rising that gives the impression that a fire has been put out. Google Translate does the best job here.
Not really. You just let the user choose which one they want for which language. 30-60 minutes of work at most.
Interesting. Have you noticed Google Translate being better for any other languages?
this guy isnt the most colorful speaker but has apparently learned an insane number of languages and addresses the issue of why the engines are better in some languages than others in this talk. i wonder if there are other resources addressing the relative strengths of each for different language–in my experience DeepL has definitely been better so I was surprised to hear this, but it probably has much less source text for asian languages
It doesn’t know Malay and keeps thinking it is Indonesian
I second this, and have just mentioned the same thing in my post about using AI. Deepl may be fantastic at other languages but it does not yet cover Malay like Google Translate can. It only does Indonesian, and after many years of making mistakes because of learning materials like that I’m sick of hearing “They’re virtually the same language”
No, NO THEY ARE NOT THE SAME!
It’s hard to trust this system now. I love the software but I cannot trust the translations