Google vice-president Vint Cerf said the use of statistical translation methods, where translations are made on the basis of probabilities and don’t rely on parsing, had vastly improved online translation.
But he warned about their reliability and said there were problems with interpreting the meaning of the same phrase in British and American English, let alone phrases in different languages.
See: The Australian
Also see: CNN
Share an article »
ProZ.com blog »
Canada
Local time: 04:55
Chinese to English
+ ...
If Google Translate really is better than all the others, then I’d say machine translation is pretty useless.
Two years ago, I happened to come across a blog entry whose author was using Google Translate (I had the suspicion, I tried to test my theory by doing a translation in Google Translate, and what I got from Google Translate matched what I saw). A sentence in English had been translated to something in Chinese with the exact opposite meaning.
So I’ve just repeated the experiment, and Google Translate is still translating my little test sentence to something that means the exact opposite of what it is supposed to mean, even though the grammatical structure in question is quite simple and common (use of “do” for emphasis).
I would say this kind of translation is pretty unreliable.
[Edited at 2011-01-26 00:11 GMT]
Netherlands
Local time: 10:55
Member (2006)
English to Afrikaans
+ ...
Heading: Google *admits*
Body text: Google *warns*
People who don't read the body text but merely glance at the headings will get an entirely different idea of what had really happened. This is worse than what happens when people get the wrong messagse when they read a machine translation instead of a human translation.
France
Local time: 10:55
English to French
+ ...
A sentence in English had been translated to something in Chinese with the exact opposite meaning.
So I’ve just repeated the experiment, and Google Translate is still translating my little test sentence to something that means the exact opposite of what it is supposed to mean, even though the grammatical structure in question is quite simple and common (use of “do” for emphasis).
I would say this kind of translation is pretty unreliable.
[Edited at 2011-01-26 00:11 GMT]
This issue happens in other languages I think, even though I do not know if this is due to the same cause as in Chinese,
In ENFR at least, I've made the test several times and GT often misses the negative forms.
Indeed the result is that the automated translation says exactly the opposite of the source.
France
Local time: 10:55
English to French
+ ...
Heading: Google *admits*
Body text: Google *warns*
People who don't read the body text but merely glance at the headings will get an entirely different idea of what had really happened.
Couldn't agree more!
United States
Local time: 04:55
Portuguese to English
+ ...
There was a great article in this month's Chronicle by Nicholas Hartmann in which he suggested the following:
"...Consider a one-paragraph medical report that has been translated into English. The concluding phrase of the translation reads: "..., the review group unanimously recommends that Mr. N not be subjected to a complete frontal lobotomy." If that entire 100-word paragraph contained only a single error involving one letter of one word, the translation could still be considered more than 99% accurate, but if the error resulted in the word "not" being changed to "now", the result would also be 100% wrong. If the person sending the translated report to the surgeon were unable to read English, that tiny but pivotal mistake would go completely undetected..."
And those of us who have experimented with GT, have seen how it sometimes misses a negative.
France
Local time: 10:55
French to German
+ ...
This issue happens in other languages I think, even though I do not know if this is due to the same cause as in Chinese,
In ENFR at least, I've made the test several times and GT often misses the negative forms.
Indeed the result is that the automated translation says exactly the opposite of the source.
The same applies to FR>DE, DE>FR and EN>DE. One should not use GT for translating legal documents, for example, or for any kind of longish sentences.
Italy
Local time: 10:55
Italian to English
This issue happens in other languages I think, even though I do not know if this is due to the same cause as in Chinese,
In ENFR at least, I've made the test several times and GT often misses the negative forms.
Indeed the result is that the automated translation says exactly the opposite of the source.
The same applies to FR>DE, DE>FR and EN>DE. One should not use GT for translating legal documents, for example, or for any kind of longish sentences.
It is the same with Italian into English. The program misses negatives, but also mangles positives. My experience with it is with the results of a survey given to Italian doctors about the quality of medical equipment used in hospitals and particularly emergency departments. Almost without exception, whenever a doctor states that a product is good at something, this is mysteriously translated as that product being bad at that thing. And when a doctor states that a product is bad, it is translated into being good. This would be amusing (i.e., this product is reliable because it breaks frequently), except that it is to do with life-saving medical equipment.
And more importantly, my mildly humorous example above is clearly nonsense, but a greater number of the reversed negatives/reversed positives do not betray the fact that they are mistranslations.
[Edited at 2011-01-26 15:49 GMT]
Canada
Local time: 04:55
Chinese to English
+ ...
I have always wondered why GT is mistranslating such common things as dates, times, indirect questions, and our friend the emphatic “do”. So at least we now know why: If GT is using statistical analysis to do the translation, then it’s clear that there’s no way it will ever be able to translate everything correctly.
[Edited at 2011-01-27 06:01 GMT]
Slovenia
Local time: 10:55
English to Slovenian
+ ...
Google Translate is so useless for translations into/from Slovene that a translator can use it as a source of entertainment while doing the real thing
I think this is excellent. It leaves that little work that we have over here to us, humans
[Edited at 2011-01-27 18:25 GMT]
[Edited at 2011-01-27 18:26 GMT]
Poland
Local time: 10:55
Member (2005)
English to Polish
+ ...
Admittedly GT has a problem with negations. I have seen it myself.
But GT is not a brain replacement. It can be used by people with brains.
PB
Thailand
Local time: 16:55
Member (2004)
English to Thai
+ ...
Many users cannot pay high fee for good translation/interpretation. Increasing roles of Google Translate to them are noteworthy. MT can be a laughing material but a very critical tool to many users in this computerized epoch. Let's seek cleverly (e.g. similar to how to use of Google Internet Search) how it can help use, the professional linguists.
Soonthon Lupkitaro
United States
Member (2009)
German to English
+ ...
I hope the people who feed KudoZ questions into Google Translate and then post completely wrong answers are reading this!
Spain
Local time: 10:55
Member (2007)
Spanish to English
I was very skeptical about GT4T and other such tools, but I gave it a chance. Now, it certainly is not going to render human translators obsolete, but it can increase productivity, and it is surprisingly good at translating simple texts. The translated text must be carefully compared to the original, and a professional translator is necessary for this task, but if it can boost productivity by even 10%, that can mean a big difference at the end of a month or year. Now, a good translator will also review to vary the style and introduce stylistic touches. But even with this task it can be worth using it. It is far from replacing the translator, but it can be a powerful tool.
Turkey
Local time: 12:55
Turkish to English
+ ...
Lost in (automatic) translation
Bosnia and Herzegovina
Local time: 10:55
German to Serbian
+ ...
I have always wondered why GT is mistranslating such common things as dates, times, indirect questions, and our friend the emphatic “do”. So at least we now know why: If GT is using statistical analysis to do the translation, then it’s clear that there’s no way it will ever be able to translate everything correctly.
[Edited at 2011-01-27 06:01 GMT]
...and generally I can't really talk about Asian languages as they have totally different language systems.
In my language, and most Euro languages, GT will do well on the level of very simple short clauses consisting only of subject and predicate ( for SVO languages). Also, it won't fail with very common words from objective reality or with common grammatical tendencies. Same goes for simple noun phrases.
However, if you have idiomatic sentence or if sentence is expanding further than this, the problems and twisted meanings begin.
Of course this doesn't mean it's 100% reliable in the former case either, it just means that it will mostly give an accurate result.
To report site rules violations or get help, contact a site moderator:
You can also contact site staff by submitting a support request »
This discussion can also be accessed via the ProZ.com forum pages.