Pages in topic: [1 2] > | How fast can DeepL translate? Thread poster: Simone De Biase
|
Dear all, I wonder if any of you know at what speed DeepL can translate. More precisely, I need to know, for academic purposes, how many words per second it can process (or words per hour, even though it is not the most appropriate value for MT software). Could you help me obtain this datum, please? Thank you all in advance! Simone
[Edited at 2019-03-09 18:35 GMT] | | | Yoana Ivanova Estonia Local time: 23:06 English to Bulgarian + ... Take a look here | Mar 11, 2019 |
http://primeurmagazine.com/weekly/AE-PR-10-17-8.html "DeepL's revolutionary neural architecture runs on a supercomputer in Iceland, capable of 5.1 petaFLOPS (5 100 000 000 000 000 operations per second), enough power to translate a million words in under a second." I can't say how reliable this information is, however. Edit: It is the information... See more http://primeurmagazine.com/weekly/AE-PR-10-17-8.html "DeepL's revolutionary neural architecture runs on a supercomputer in Iceland, capable of 5.1 petaFLOPS (5 100 000 000 000 000 operations per second), enough power to translate a million words in under a second." I can't say how reliable this information is, however. Edit: It is the information given everywhere online, where the speed is mentioned.
[Edited at 2019-03-11 09:04 GMT] ▲ Collapse | | | Samuel Murray Netherlands Local time: 22:06 Member (2006) English to Afrikaans + ... Unknown, really | Mar 11, 2019 |
Yoana Ivanova wrote: "DeepL's revolutionary neural architecture runs on a supercomputer in Iceland, capable of 5.1 petaFLOPS (5 100 000 000 000 000 operations per second), enough power to translate a million words in under a second." This information is from a press release (what would explain why it is repeated everywhere). The "5.1" figure is sufficiently precise to make it believable, but "a million" sounds like a rounded-off number. Besides, the fact that DeepL runs on this computer doesn't mean the entire supercomputer is available to DeepL. It is doubtful that the DeepL company actually owns the supercomputer. In addition, the speed of translation would be limited to the speed of input (and possibly output as well). | | | Tom in London United Kingdom Local time: 21:06 Member (2008) Italian to English
I imagine it would also depend on how fast your computer or device is, the quality of your router and how you are connected to it, the quality of your ISP and, crucially, how far away you are from the main servers (in Iceland, apparently).
[Edited at 2019-03-11 09:23 GMT] | |
|
|
Simone De Biase Italy Local time: 22:06 English to Italian + ... TOPIC STARTER
Thank you kindly, Samuel, Yoana, and Tom!
[Edited at 2019-03-11 09:40 GMT] | | | Yoana Ivanova Estonia Local time: 23:06 English to Bulgarian + ... Theoretical number | Mar 11, 2019 |
Samuel Murray wrote: It is doubtful that the DeepL company actually owns the supercomputer. In addition, the speed of translation would be limited to the speed of input (and possibly output as well). I agree, with a bit more googling one can find that the supercomputer is owned by Verne Global. It's more of a theoretical number, if the full capicity is employed, which it likely never will be. But it doesn't seem like there's any better information available to the public out there. | | | Tom in London United Kingdom Local time: 21:06 Member (2008) Italian to English Fast doesn't mean it's good | Mar 11, 2019 |
No matter how fast it is it still produces laughably inaccurate translations. For a good translation you need to go through the source text slowly and carefully - not as fast as you can. And then you need to go over your first translation again, correcting it. And then you need to do it again, and probably two or three more times. Then you need to sleep on it, and read it again, maybe even printing it and correcting the printed copy before finalising it. Deepl can't do that becau... See more No matter how fast it is it still produces laughably inaccurate translations. For a good translation you need to go through the source text slowly and carefully - not as fast as you can. And then you need to go over your first translation again, correcting it. And then you need to do it again, and probably two or three more times. Then you need to sleep on it, and read it again, maybe even printing it and correcting the printed copy before finalising it. Deepl can't do that because a good translation requires slow, attentive, critical evaluation.
[Edited at 2019-03-11 12:55 GMT] ▲ Collapse | | | Simone De Biase Italy Local time: 22:06 English to Italian + ... TOPIC STARTER Thank you kindly, Yoana | Mar 11, 2019 |
Yoana Ivanova wrote: http://primeurmagazine.com/weekly/AE-PR-10-17-8.html "DeepL's revolutionary neural architecture runs on a supercomputer in Iceland, capable of 5.1 petaFLOPS (5 100 000 000 000 000 operations per second), enough power to translate a million words in under a second." I can't say how reliable this information is, however. Edit: It is the information given everywhere online, where the speed is mentioned. [Edited at 2019-03-11 09:04 GMT] Thank you kindly, Yoana! I'll ask my co-supervisor how I could use that value. P.S.: I've quoted you on my MA thesis that I'm currently writing, Yoana! | |
|
|
Yoana Ivanova Estonia Local time: 23:06 English to Bulgarian + ... Couldn't agree more | Mar 11, 2019 |
Tom in London wrote: No matter how fast it is it still produces laughably inaccurate translations. You're absolutely right. It'll be a long time before MT can really claim to produce output remotely comparable to that of a human translator. Although DeepL in particular has made such claims. I did a comparison once just to get a good laugh out of it. Both DeepL and Google Translate are terrible, but DeepL with all its claims had the worse result. However, the OP said it's for academic purposes, so I don't think he will actually be using it for translation, just wants to have the numbers. | | | Yoana Ivanova Estonia Local time: 23:06 English to Bulgarian + ... | Tom in London United Kingdom Local time: 21:06 Member (2008) Italian to English
Oh dear....I used to rent an apt. from Cushman and Wakefield. "Dear landlord Please don't put a price on my soul" (Bob Dylan)
[Edited at 2019-03-11 10:32 GMT] | | | Simone De Biase Italy Local time: 22:06 English to Italian + ... TOPIC STARTER Yoana is right, I just need the numbers. | Mar 11, 2019 |
Yoana Ivanova wrote: However, the OP said it's for academic purposes, so I don't think he will actually be using it for translation, just wants to have the numbers. Indeed. I have done a translation experiment for my MA thesis that I'm currently writing and hope to finish within the end of next week (fingers crossed) Wish me good luck, please!
[Edited at 2019-03-11 11:24 GMT] | |
|
|
Simone De Biase Italy Local time: 22:06 English to Italian + ... TOPIC STARTER Thank you, Yoana | Mar 11, 2019 |
Thank you very much, Yoana. I hope I'll have the time to give a look at those links. Thank you, again! | | | Yoana Ivanova Estonia Local time: 23:06 English to Bulgarian + ...
Simone De Biase wrote: Indeed. I have done a translation experiment for my MA thesis that I'm currently writing and hope to finish within the end of next week (fingers crossed) Wish me good luck, please!
[Edited at 2019-03-11 11:24 GMT] Good luck with your thesis! I'd be interested in reading it when you're finished. If that's okay, I'd be happy, if you drop me a message/email. | | | Fast and bad quality | Mar 12, 2019 |
never accept post-editing of MT-translations. Any translator should refuse to handle this kind of task. Translation is a complex art, translators build the bridges between countries, beautiful bridges and not fast no issue highways. And by the way, handle these kind of translation is like cutting the branch on which we translators are sitting ! | | | Pages in topic: [1 2] > | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » How fast can DeepL translate? Trados Studio 2022 Freelance | The leading translation software used by over 270,000 translators.
Designed with your feedback in mind, Trados Studio 2022 delivers an unrivalled, powerful desktop
and cloud solution, empowering you to work in the most efficient and cost-effective way.
More info » |
| Anycount & Translation Office 3000 | Translation Office 3000
Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |