Pages in topic:   < [1 2 3 4 5 6 7 8 9] >
DeepL
Thread poster: Michael Mangelsdorf
Giovanni Guarnieri MITI, MIL
Giovanni Guarnieri MITI, MIL  Identity Verified
United Kingdom
Local time: 05:30
Member (2004)
English to Italian
Absolutely... Aug 30, 2017

CafeTran Training wrote:

Let's wait for the API. I'm eager to have this new MT system integrated in my favourite CAT tool .


Without it, it's pretty much useless...


 
Tom in London
Tom in London
United Kingdom
Local time: 05:30
Member (2008)
Italian to English
I played with it too Aug 30, 2017

Giovanni Guarnieri MITI, MIL wrote:

EN > IT... I don't think it's even better than GT right now...


I played with it too and it's just as bad as GT. A case of "so near, but yet so far". Machine translation does not recognise mistakes in a badly written source text, or makes corrections of words that did not require correction, e.g. reading "essenza" as "assenza" and offering "absence" as the translation - which is way off.

And so on. Translators can continue to sleep easy in their beds.

HINT: do not accept DeepL's invitation to "provide a better word". Unless you want to be digging your own grave. DeepL relies on ingenuous translators feeding it with corrected texts. On the other hand, you could sabotage it by suggesting completely wrong words....

[Edited at 2017-08-30 12:18 GMT]


tabor
Yaotl Altan
 
Giovanni Guarnieri MITI, MIL
Giovanni Guarnieri MITI, MIL  Identity Verified
United Kingdom
Local time: 05:30
Member (2004)
English to Italian
Agreed... Aug 30, 2017

Tom in London wrote:

Translators can continue to sleep easy in their beds.


If this is the future... also, the Italian texts on Linguee are not something I would trust... they are always badly translated, even the EU's ones, sometimes...


 
Tom in London
Tom in London
United Kingdom
Local time: 05:30
Member (2008)
Italian to English
Agreed Aug 30, 2017

Giovanni Guarnieri MITI, MIL wrote:

Tom in London wrote:

Translators can continue to sleep easy in their beds.


If this is the future... also, the Italian texts on Linguee are not something I would trust... they are always badly translated, even the EU's ones, sometimes...


Agreed - you have to be careful because many of the "English translations" offered by Linguee actually originate from Italian sources and are full of mistakes.

The fact is that those who are spending $millions on trying to come up with the perfect, or near-perfect MT system do not actually understand anything about languages.

[Edited at 2017-08-30 12:28 GMT]


 
Natasha Ziada (X)
Natasha Ziada (X)  Identity Verified
Australia
Local time: 15:30
English to Dutch
+ ...
Worryingly good Aug 30, 2017

I tested DeepL for English - Dutch, some legal bits and pieces, and I honestly could hardly find any errors, or issues with readability, at all. More natural sounding than GT, at least in that area/language pair. It actually worries me - apart from the issue of confidentiality, what is stopping potential clients from using technology that is both fast and free, rather than a professional?

Nikolay Nikolov
 
Timothy Barton
Timothy Barton
Local time: 07:30
French to English
+ ...
Are you sure they understand nothing? Aug 30, 2017

Tom in London wrote:
The fact is that those who are spending $millions on trying to come up with the perfect, or near-perfect MT system do not actually understand anything about languages.

[Edited at 2017-08-30 12:28 GMT]


Are you sure they understand nothing? I know people who have worked in the sector, and they are very knowledgeable, which is why MT is so much better nowadays than it was in the days of rules-based MT. Just because MT can't perform like humans doesn't mean the designers don't know anything about languages. I think most are aware that MT will never replace humans, but that doesn't stop them trying to make it as good as they can.


Matheus Chaud
 
CafeTran Training (X)
CafeTran Training (X)
Netherlands
Local time: 06:30
Slator Aug 30, 2017

See also: https://slator.com/technology/linguees-founder-launches-deepl-attempt-challenge-google-translate/

 
Tom in London
Tom in London
United Kingdom
Local time: 05:30
Member (2008)
Italian to English
What's stopping them Aug 30, 2017

Natasha Ziada wrote:

....what is stopping potential clients from using technology that is both fast and free, rather than a professional?


What's stopping them is the need for top-drawer 100% correct translations done by a sentient being who understands all the vocabulary, is able to detect errors, incorrect punctuation, misspellings, poor sentence construction, etc. in the source text, and who perceives nuance, ambiguity, irony, and style. Show me an MT algorhythm with the ability to do that, to translate really important documents (e.g. an Airbus A380 flight manual, a manual for the removal of a brain abcess, a scholarly assessment of the sculpture of Gian Lorenzo Bernini, a critical essay on aesthetic theory, or a badly written bill of quantities for alterations to an historic building) without making any mistakes or missing any allusions, and I'll retire from translating.

[Edited at 2017-08-30 16:05 GMT]


Mijndert Hermes
Yaotl Altan
Francisco González Ramírez
 
CafeTran Training (X)
CafeTran Training (X)
Netherlands
Local time: 06:30
English as an interlingua? Aug 30, 2017

Has it already been established whether English is an interlingua for DeepL too?

 
Vanda Nissen
Vanda Nissen  Identity Verified
Australia
Local time: 15:30
Member (2008)
English to Russian
+ ...
Not there yet Aug 30, 2017

I've tested it for Polish, just a random article, and it is not impressive but there are more differences between English and Polish in terms of the grammar, than, for example, between English and Dutch, so I am not surprised.

I do not agree with Tom either - there are a lot of linguists, mostly computational linguists, who are working on the MT development. They are highly qualified specialists.


 
Kay-Viktor Stegemann
Kay-Viktor Stegemann
Germany
Local time: 06:30
English to German
In memoriam
Better than GT but not really impressive Aug 30, 2017

The first things I tried with it looked really impressive, but on a closer look it fell apart soon. I'd agree that it is better than GT, because it seems that it is able to produce a fluent sentence structure in German more often. And it could be useful with an API or to get the basics of a text in an unknown language, but it is still far from being usable for real translation.

The interesting thing is that it makes other mistakes than GT, for example I found that it leaves out part
... See more
The first things I tried with it looked really impressive, but on a closer look it fell apart soon. I'd agree that it is better than GT, because it seems that it is able to produce a fluent sentence structure in German more often. And it could be useful with an API or to get the basics of a text in an unknown language, but it is still far from being usable for real translation.

The interesting thing is that it makes other mistakes than GT, for example I found that it leaves out parts of the source in some cases. I believe that this neural network learns from previous translations and imitates them. It is kind of a parrot, it is able to reproduce translations it knows and even to recombine them. It might be a very knowledgeable parrot with terabytes of language data in its parrot brain, it might be able to form sentences but it still does not know what it says.
Collapse


 
Michael Beijer
Michael Beijer  Identity Verified
United Kingdom
Local time: 05:30
Member (2009)
Dutch to English
+ ...
ha ha, Aug 30, 2017

Kay-Viktor Stegemann wrote:

The first things I tried with it looked really impressive, but on a closer look it fell apart soon. I'd agree that it is better than GT, because it seems that it is able to produce a fluent sentence structure in German more often. And it could be useful with an API or to get the basics of a text in an unknown language, but it is still far from being usable for real translation.

The interesting thing is that it makes other mistakes than GT, for example I found that it leaves out parts of the source in some cases. I believe that this neural network learns from previous translations and imitates them. It is kind of a parrot, it is able to reproduce translations it knows and even to recombine them. It might be a very knowledgeable parrot with terabytes of language data in its parrot brain, it might be able to form sentences but it still does not know what it says.


… that sounds like me!

Michael


 
Claudio Porcellana (X)
Claudio Porcellana (X)  Identity Verified
Italy
Local time: 06:30
English to Italian
yes, very good for certain language couples and topics, but ... Aug 30, 2017

my opinion is that this tool is very reliable to find "pieces of a translation", less commonly a whole good translation, but a thinking brain is still needed (luckily enough) to properly join these pieces together, so I think we will not replaced in the short by HAL 9000


about the privacy I still not understand the point: unless you translate patents, a new product just before its launch on the market, or some s
... See more
my opinion is that this tool is very reliable to find "pieces of a translation", less commonly a whole good translation, but a thinking brain is still needed (luckily enough) to properly join these pieces together, so I think we will not replaced in the short by HAL 9000


about the privacy I still not understand the point: unless you translate patents, a new product just before its launch on the market, or some secret military paper, the files received to translate fall into the public domain, so you breach nothing taking the wraps off your docs to GT, Lingue or even the local café


a last note: these system rely on our (yours) translations, so the "average quality" cannot be higher than the "average quality of the average translator" that is not so high IMHO: the logical consequence is that the above average translator has nothing to fear from these tools

this is because these systems rely on collecting large quantity of data, i.e. no Artifical Intelligence is involved here

the day a IT system will be able to "reason, speculate, solve novel problems" with its cybernetic brain, well this will be quite another story

my 2 cents

[Edited at 2017-08-30 23:13 GMT]
Collapse


 
Claudio Porcellana (X)
Claudio Porcellana (X)  Identity Verified
Italy
Local time: 06:30
English to Italian
yes, very good for certain language couples and topics, but ... Aug 30, 2017

sorry double post

[Edited at 2017-08-30 23:11 GMT]


 
CafeTran Training (X)
CafeTran Training (X)
Netherlands
Local time: 06:30
Searches are case-sensitive Aug 31, 2017

I just noticed that searches are case-sensitive:

Screen Shot 2017-08-31 at 09.56.10

And:

Screen Shot 2017-08-31 at 09.56.18

Can anyone find a URL that includes the language settings for the SL and TL? I can see translatorSelectGerman in the source but I don't know how to add it to the URL.


 
Pages in topic:   < [1 2 3 4 5 6 7 8 9] >


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

DeepL






Wordfast Pro
Translation Memory Software for Any Platform

Exclusive discount for ProZ.com users! Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value

Buy now! »
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 »