I just got back from the latest Copeland Colloquium at Amherst College, “Translation in the Age of Google Translate,” with languages and literature faculty members Cathy Ciepiela, Laure Katsaros, and Andrew Parker. It was convenient (it was in the Frost Library, which I was familiar with from my afternoon with Cowan, and they ended it just before one so people could get to classes, which meant I could leave to meet my wife without missing any of the discussion) and quite interesting, but I was a bit annoyed by the tone of it, which was largely negative: not only are the results of Google translation unreliable, but it’s done (shudder) for profit! To illustrate how bad it was, one of the panelists quoted a passage from Flaubert as translated by Google: ha, Google used “it” instead of “she”! During the discussion period, I pointed out that people, by and large, do not use Google Translate on Flaubert but on more mundane documents like news stories, and they are not so much concerned about grammatical perfection as about getting the gist, at which Google Translate is usually pretty good. Another audience member pointed out that many people around the world do not have the kind of access to human translators and translation services that “we at Amherst” have, and Google Translate is their only means of accessing the world outside of their own language. The panelists were gracious enough to concede these points and admit that they were, inevitably, looking at it from their own perspective as translators and language teachers. It was an enjoyable way to spend the lunch hour (and the lunch was excellent!); I expect I’ll be going to other such talks.
Well said, Language! I use Google translate all the time, I love it. It’s not supposed to be very accurate into Spanish, according to Julia, but it’s quite good for Norwegian.
I’m missing something here. How does Google generate profit from Google Translate, and how does that differ from a translation agency’s profit or a translator’s fee?
Google’s weakness in the Spanish-to-English realm surprises me. But I don’t know how it compares to other subject-pronoun-optional languages.
The Google Translate web page is not monetized, but the Google Translate API is. It’s for web sites that want to auto-translate their continually varying content.
As I have said before (perhaps even here), Google works like a non-profit for the most part, except that when a worthy project needs more money, it goes to the money pit in Google’s basement (labeled “Adwords/Adsense”) and dips out some more money. You’d think the advertising teams would be the most important ones internally, but they aren’t.
I’m missing something here. How does Google generate profit from Google Translate, and how does that differ from a translation agency’s profit or a translator’s fee?
You’re getting a little too detailed. Google is a for-profit company, part of the capitalist snake that is strangling humanity (to put lurid words in the mouth of the mild-mannered professors, who would doubtless never express themselves thus, but whose way of looking at the world I think I have captured). They themselves, being professors, are used to being underpaid for translating, as for scholarly work in general, so they probably don’t see the parallel.
I reckoned they were entitled to the benefit of the doubt, but expected them to be as you described.
I will come out and admit that I have got very lazy. I use Google Translate when I could translate things myself. Why? Because it removes a lot of the drudgery of translation. Translating complex sentences can be quite exhausting; with Google translate you just need a bit (well, sometimes a lot) of rearranging of already-translated material. Translating specialised terms, set phrases, names, etc. often means tedious searching; Google translate can be very good at digging up the right term or name, and if it gets it wrong you can usually tell. Of course it’s useless for translating literature, but for more mundane translation it is great for getting some of the mundane legwork out of the way.
I’m also wondering if Google translate isn’t subtly changing the results of translation. Good translators know how to recast sentences in creative and flexible ways. Bad translators slavishly recreate the original structure. But Google translate shakes sentences up in strange ways that could force bad translators to abandon rigid adherence to the original. It’s a loosener of sorts. It probably means the demise of the well-crafted sentence as translators (by default) forfeit their creativeness to Google. Whether this is a good thing or a bad thing, I am unable to judge.
Robe, are you serious? About the effect on bad translators I can imagine that you are right, but what about good translators?
@ Zero. Muddled musings. But I’m talking here partly about the culture of translation. For example, quite a lot of the nuts and bolts of translation in languages like Japanese and Chinese have been built on English grammar as learnt in school. That’s how translators learn to manipulate sentences, and even good translators are unable to totally escape from it. As people rely more on tools like Google translate, will the habits and work-arounds change? Perhaps this won’t apply in English so much, as modern English is possibly less prone to be influenced by translations.
I’m also wondering if Google translate isn’t subtly changing the results of translation.
You bet it is!
That’s because when it translates a text, it looks not only at previously existing parallel texts in the selected language pair, it also looks at the results of its own previous translations.
It’s like traveling in circles of ever-decreasing radius. And you know where that leads.
Well, if they are concentric circles of ever-decreasing radius then I know where it leads. Non-concentric circles, I’m not saying.
I’m afraid I have just as negative a view of Google Translate and it’s cousins like Youdao Translate, but for slightly different reasons. Over the last few years I’ve had more and more students handing in essays that were very clearly and obviously written first in Chinese (by them or somebody else) then run through Google or Youdao Translate. Every time I hand it back and tell them to do it properly and give me a real essay next class. Last semester I had one student give me an essay. I glanced at it and instantly recognised the format. “This is a Baidu Baike article, isn’t it?” “Yes.” “And you used Youdao Translate?” “Yes.” “Go away and write it yourself, give me a real essay next lesson.” She comes back with a new ‘essay’. Another Baidu Baike article run through Youdao Translate. Zero. How these kids expect to learn to write English essays, I don’t know. And this all despite me demonstrating in class just how useless translations software is* and warning them that I will tell at a glance whether they’ve used translation software or plagiarised.
*Alright, as Bathrobe points out, it is quite good for technical terms and formulaic language like addresses (It can render Chinese addresses into good, if backwards, English down to the county/district level. Anything below that (street/housing estate/etc) you have to translate yourself), but those are areas where you do actually have direct one-to-one correlations between words. I think the whole translation software project is fundamentally misguided because it works on the assumption that the words actually have any meaning or significance in and of themselves. They do not. They are mere vehicles for ideas. Using translation software for anything beyond those technical/formulaic areas is as fundamentally dumb as writing an anthropological study of a city’s people based purely on what cars are driven on that city’s streets.
Paul Ogden: Do you have evidence for this? When I worked at Google (but not on GT), I was assured otherwise. Human corrections entered into the web page are taken into account, but not automatically, which is why you don’t see instant results.
On some news blogs, I get the impression that the use of Google Translate is a kind of guarantee of impartiality in the translation. The nuances may be scrambled, but that’s better than the nuances being affected by the blogger’s own biases. I’m not sure where I got this impression from, but on at least one occasion I thought a blogger who provided his own translation of a foreign news article come across as somewhat defensive about it, as if he had to justify not using Google Translate.
Ken
@John Cowan re Paul Ogdan’s comment. There was a bit of concern about a year ago about this issue. Statistical translation depends on having a large corpus of high-quality translated material as input, and the current translated material has more and more machine-translated material polluting the input stream.
How, or even whether, Google is addressing this is something I don’t know.
As (sadly) a single language person, without Oogle translate, I would be lost in any non-English language site. It’s important to me to get the general idea. It’s always clear it’s not a real translation, just a sense of what is going on.
Plus, there is the amusement. This sentence from an entry on a folk painted Lion-house.
“The house is uninhabitable until the redemption of his project team was under threat of dismantling its owners. ”
Comedy gold.
John Cowan: I recall reading something about this a while back but cannot find it now.
My search did come up with an amusing NYT item, Putting Google to the Test in Translation.
The article compares human and Google translations of several short bits from well-known literary works.
@Paul Ogden, John Roth: I think you’re thinking of this old Language Hat entry and the Atlantic article it links to. Note that the problem in question does not affect Google Translate.
Bathrobe, I’m afraid I don’t agree with you much. Bad translations come in a great variety of badness. It’s not just slavish adherence to the original syntax. Good translators don’t use Google Translate (although they often use other translation programs). So I don’t think Google or other online translation programs have much effect on translation as a professional endeavor.
That said, bad translations do have a great effect — not on translators, but on clients. They get used to them and begin to think that they are perfectly fine, that the oddities of syntax and word choice is appropriate quality of translation, that a translated text should sound weird. In Russia, where about 90 percent of the translations into English are done by Russian-speakers, clients expect a certain kind of bad translation and criticize a good one, ie, one done by a native speaker. I can’t tell how many times some punk kid has returned my translation as “not done by a native speaker” and marking my use of articles (!) or word choice (insisting, for example, that “philanthropy” can’t be used to describe a foundation’s “social programs”).
I’m with mab, the real damage that something that google translate does is degradation of standards. If a person has a clear idea of what a program can and cannot do, then it’s a valuable service. Most people have no such idea. And the last thing that working translators need is something else that makes potential clients think they can be replaced by a free web service.
And I’ve had similar experiences in Poland. No Polish person has ever claimed I’m not a native speaker but I’ve been challenged about a lot of issues because teaching practices often contradict native usage and/or typical translation practice.
I was once criticized for not using British verb agreement and many of my explanations about word choice and/or sentence structure are met with barely concealed skepticism.
I was indirectly challenged by a German (who I never met) who apparently claimed that the changes I introduced into an English text (written by a Pole) couldn’t have been done by a native.
Norwegians & Germans have occasionally corrected or been skeptical about my use of English. I find such lack of deference very irritating.
I thought the snowclone was “X in the age of mechanical reproduction”; I should have realized it was general enough to create “X in the age of Y”.
Norwegians & Germans have occasionally corrected or been skeptical about my use of English. I find such lack of deference very irritating.
Same here, for Germans (I guess I haven’t hung out with enough Norwegians). Seriously, I’m a native speaker and what I say goes, I don’t care what your memory of your school grammar book says.
Ran wrote: Note that the problem in question does not affect Google Translate.
It seems to me that it does. If enough material rendered into another language by Google’s Translate API finds its way into Google’s corpus, then the corpus becomes compromised. By the same token, any text on any website that’s been created by plain ol’ vanilla Google Translate could also find its way into the corpus.
Norwegians & Germans, etc: Native Hebrew speakers occasionally try to instruct me in the correct use of English. Drives me right up the wall, especially when their original Hebrew text is poorly written. Arrgh!
Paul: But as I keep saying, random translated web pages don’t get put in Google’s translation corpus. Googlers, whatever their faults, simply aren’t dumb in that way.
It’s odd to me, the ardent belief that one’s grammar book is Perfect. When Russians use phrases “prohibited” in my grammar books (or change stress, which happens a lot), I always immediately assume that my grammar book was wrong or is now outdated. But Russians nearly always assume that I’m wrong. Or that I’m speaking a very sub-standard variety of — shudder — American English. Once during a presentation I said, “To my mind…” The entire hall started giggling. Finally a professor explained that their grammar books insisted that the phrase “to my mind” was hopelessly outdated and bookish. In the end, I left with the impression that the entire hall had dismissed my speech as bookish rather than re-evaluate their grammar book.
Weirdest example (from when I still lived in a teachers dorm). I get a knock on the door and a another resident as in heavily Russian accented Polish if I can help her friend on a ‘business English’ test. She has a list of terms she needs to know and hands me a handwritten list that looks something like this:
iaaajgh
versytuio
gutyyt sis!ios
vysu#ess buek
(I just made up those, but it mostly looked like random sequences of letters with a few indecypherable symbols thrown in for good luck).
I said I was sorry but I don’t recognize any of them as English words. She looked at me skeptically and asked if I was actually an English teacher. I assured her again and apologized for not being able to help.
“Maybe these are newer words” she said sympathetically nodding as if I were a poor fraud who she felt too sorry for to expose.
To be fair, many times non-native speakers have made interesting and insightful comments on usage (that I hadn’t noticed).
Also when I was teaching translation classes I lost track of the times my students came up with better wording of particular sentences (or parts of sentences) than I had.
But it still rankled when people began challenged points of usage based on what they learned 15 years ago or a sentence they found on the internet…
It occurs to me that that’s basically the same phenomenon as peevery in one’s own language: refusal to listen to experts (linguists in one case, native speakers in the other) and clinging to bits of “knowledge” picked up from sources one considers authoritative.
In the spirit of fairness — native speakers often disagree, too. (But native speakers usually defer to each other; there are plenty of usages that I consider “dated and bookish”, but that I wouldn’t blink an eye at if I heard a fellow native speaker use them unselfconsciously in straightforward speech. I think that non-native listeners often pay more active attention to the syntax of what’s being said, so they consciously notice and care about things that a native listener would not.)
John Cowan: But as I keep saying, random translated web pages don’t get put in Google’s translation corpus. Googlers, whatever their faults, simply aren’t dumb in that way.
If I take a text in German, ask Google Translate to do its magic and render it into Spanish, then place the result in a document that is eventually placed on a website, it seems to me that the final text could indeed find its way into Google’s corpus.
Or perhaps I’m missing something.
Norwegians & Germans have occasionally corrected or been skeptical about my use of English. … Seriously, I’m a native speaker and what I say goes, I don’t care what your memory of your school grammar book says.
I couldn’t count the times one or the other whippersnapper of a German programmer has tried to correct me as to the best way to formulate something in English. My everyday German is perfectly idiomatic, and my formal prose is far superior to what is produced by run-of-the-mill educated Germans – but I do have a slight American accent (oh the shame of it!). If these know-it-alls have ears to hear they could not seriously imagine they are arguing about English on an equal footing with another German person.
This phenomenon really puzzles me. It does provide further proof that society consists of monads talking to themselves (this is one of Habermas’s caricatures of Luhmann’s ideas). But Luhman never said monads can’t get nasty, so I do. What I say to the know-it-alls nowadays is direct and brutal: “You are in no position to correct me. English is my mother tongue, German is yours. If you think you know English, let’s start talking in English and see how far you get.”
The effect of this little speech is also strange: they look at me like a dog that doesn’t understand why you kicked him. No enlightenment, just puzzled doggy eyes.
Haha. I’ll try that.
Or perhaps I’m missing something.
What you’re missing, though I don’t know how after all these times, is that Google’s translation team simply does not use random pages off the Web to seed its corpora. It uses bilingual texts developed by human translators.
“It’s like traveling in circles of ever-decreasing radius. And you know where that leads.”
Yeah, right up its own arsehole.
Translators’ obsession with google translate is a smokescreen to hide the very real advantages brought to the translator by using google search.
Having access to such a huge multilingual corpora of parallel texts is MASSIVE. But, to quote Ivor Cutler, nobody talk about it.
Google image search in particular is very helpful for certain thorny parts of translation-work.