Southwest path information gap, a cloister

« previous post | next post »

Someone recently told Tom Bishop (creator of Wenlin software for learning Chinese) that Google Translate is really good now, so he tried translating this English paragraph into Chinese (chosen randomly from the cave adventure game):

You are on one side of a large, deep chasm.  A heavy white mist rising up from below obscures all view of the far side.  A southwest path leads away from the chasm into a winding corridor.

The result is:

Nín shì yīgè dà de, shēn de hónggōu de yībiān. Yīgè chénzhòng de bái wù, cóng xiàmiàn shàngshēng yǎngàile suǒyǒu de yuǎnfāng de kànfǎ. Xīnán lùjìng xìnxī hónggōu, chéngwéi yīgè huíláng. 您是一个大的,深的鸿沟的一边。一个沉重的白雾,从下面上升掩盖了所有的远方的看法。西南路径信息鸿沟,成为一个回廊。

And the result of converting that Chinese back to English is:

You are a big, deep side of the gap. A heavy white mist, rising from below to cover all of the distant view. Southwest path information gap, a cloister.

Converting that back into Chinese produces:

Nǐ shì yīgè dà de chājù shēn chù. Yīgè chénzhòng de bái wù, cóng xiàmiàn shàngshēng dào hángài suǒyǒu de yuǎnjǐng. Xīnán lùjìng xìnxī de chājù, huíláng. 你是一个大的差距深处。一个沉重的白雾,从下面上升到涵盖所有的远景。西南路径信息的差距,回廊。

Back to English:

You are a big gap between the depths. A heavy white mist, rising from below to cover all of the vision. Southwest path information gaps, corridors.

Back to Chinese:

Nǐ shì yīgè hěn dà de chājù zhī jiān de shēndù. Yīgè chénzhòng de bái wù, cóng xiàmiàn de shàngshēng, yǐ hángài suǒyǒu de yǎnguāng. Xīnán lùjìng xìnxī chājù, zǒuláng. 你是一个很大的差距之间的深度。一个沉重的白雾,从下面的上升,以涵盖所有的眼光。西南路径信息差距,走廊。

English:

You are a big gap between the depth. A heavy white mist, rising from below, to cover all of the vision. Southwest path information gap, the corridor.

My intention in making this post is not to disparage Google Translate, because I actually think that it is not bad, but to recognize that it performs as well as it does and to caution that one can never rely on machine translation alone, but must consult someone who is adept in both source language and target language.  Moreover, Google Translate has other virtues beyond providing a rough-and-ready initial translation for those with less than full competence in the source language.  As Cindy Ning says, "I now use Google Translate almost as often as I use Wenlin! Not for translation (as Tom's funny exercise above warns against), but as a lookup tool for vocab. And to produce pretty good pinyin transcriptions!"



49 Comments

  1. Felix said,

    January 19, 2012 @ 12:50 am

    This is amusing. I am wondering, have human translators been subject to study of their translation abilities in this way? That is, back translating without access to the original material? Curious how they'd fare in such a study.

  2. Jim said,

    January 19, 2012 @ 1:11 am

    Not really the best paragraph to test Google Translate with. Next time he should try a simpler one.

  3. Bears said,

    January 19, 2012 @ 1:11 am

    @Felix

    That would be a most interesting game of "broken telephone!"

  4. Honey said,

    January 19, 2012 @ 1:24 am

    Am I alone in thinking the final translation displays a kind of ascetic beauty? Not being a Chinese speaker it makes me wonder if this is accident or some English-speaking cultural interpretation of Chinese sentence form.

  5. DG said,

    January 19, 2012 @ 2:41 am

    It's a well-known game, I think, to have a bunch of people who speak multiple languages, e.g. A speaks X and Y, B speaks Y and Z, C speaks Z and W and so on, and start with some text in language X, passing it through the sequence, ideally ending back in X. The results are usually very funny. Probably you can find some examples if you google it.

  6. LDavidH said,

    January 19, 2012 @ 3:24 am

    When I was working in Albania, we often had Albanians translate various texts we needed in our teaching. Unfortunately, most of them were very wooden in their translations, using unnatural expressions and word-for-word translations. So it fell to me, a fluent but still non-native speaker of Albanian, to read through every translation and "correct" it (sometimes by asking, "Is this really how you would say it?"; sometimes by explaining what the English actually meant). Not quite the same thing as Felix is suggesting (less amusing, for one thing!), but still illustrating the need for doublechecking translations!

  7. Tom Saylor said,

    January 19, 2012 @ 5:16 am

    You'd think that the Google translator would have a rule specifying that the object of 'between' must be plural (regardless of the number of the corresponding noun phrase in the non-English text) and that an output like "between the depth" would be impossible.

  8. Brindi said,

    January 19, 2012 @ 6:16 am

    @ LDavidH
    I am a "native" Albanian speaker (was born there, but learned the language from family members while abroad) and am always very impressed when I come across the rare learner of the language, as I still find it almost impossible to speak without making any mistakes. Family members have often asked me to translate various phrases and texts from movies, newspapers, magazines, etc, and I have always found it much harder to translate Albanian expressions into English rather than English into Albanian, even though I know English much better than Albanian. It would be fun to pass around a piece of text, as Felix suggests, among friends and family who speak both languages to see what they come up with.

  9. P said,

    January 19, 2012 @ 6:42 am

    Google translate is still pretty bad for Hungarian, but it has gotten better. Before, it didn't know which words to put the case endings onto, so it would write the case endings as separate words, for example -ra for "onto", and if there was more than one plausible noun in the sentence it was hard to tell which noun the case ending was supposed to be attached to. I just tried translating the english paragraph you supplied and it did a passable job, not fully intelligible but at least it made some use of cases.

    Ön az egyik oldalon egy nagy, mély szakadék. A nagy fehér köd emelkedett alulról eltakarja az összes képet a túlsó. A délnyugati út vezet el a szakadék egy kanyargós folyosón.

    case endings:
    oldalon = on the side (i wouldn't use oldál in this sentence though, I would say "Ön egy nagy, mély szakadék előtt áll" (You a large, deep chasm in-front-of stand).

    alulról = from underneath
    képet = picture (accusative)

    But, the verb is wrong in this sentence, inexplicably rendered in the past tense. It would be better to say "Az alulról emelkedő nagy, fehér ködtől nem látszik a túl oldal." The from-underneath rising large white fog-[lit. away-from fig. because of] not visible the far side.

    folyosón – on a cooridor. Wrong case ending but at least it tried. The sentence is somewhat intelligible if incorrect.

    Translating back into English, we get:

    You are on one side of a large, deep ravine. A white mist rose from the bottom of the picture is covering all of the other. The road to the south-west divide is a winding corridor.

    and back into Hungarian:

    Ön az egyik oldalon egy nagy, mély szurdok. A fehér köd emelkedett alulról a kép, amely az összes többi. Az út a dél-nyugati megosztottság kanyargós folyosón.

    Back into English:

    You are on one side of a large, deep ravine. A white mist rose from the bottom of the picture that all the rest. The road to the south-west divide is winding corridor.

    Back into Hungarian:

    Ön az egyik oldalon egy nagy, mély szurdok. A fehér köd emelkedett alulról a kép, hogy az összes többit. Az út a dél-nyugati megosztottság kanyargós folyosón.

    and back into English:

    You are on one side of a large, deep ravine. A white mist rose from the bottom of the picture that all the rest. The road to the south-west divide is winding corridor.

    The English still looks vaguely good but the Hungarian is now completely incomprehensible.

    But let's give it a try with an easier text, one from a children's book. Google translate gives us:

    Winnie the Pooh One day, when there was nothing to do, remembering that I should make something very important. He went so Malackához to explain what it does exist, but Malackánál just could not find anyone. So I went back home while the snow fell heavily and thought that maybe at home, there's a little food. How large are good and kimelegedjék jumped into the cold and the light began to sing.

    This should of course be:

    One day, when Winnie the Pooh did not have much to do, it occured to him that he should do something important. Therefore, he went to Piglet's house, to see what he was doing, but at the moment, he found there was nobody at home. So he headed homewards, while the snow fell thickly, and he thought that perhaps at home there might be a little something to eat. To warm up, he jumped about and took giant steps, and with respect to the cold he began to sing.

  10. richard howland-bolton said,

    January 19, 2012 @ 7:17 am

    I tried the back and forth experiment back (if not forth) in 2003.
    http://howlandbolton.com/essays/read_more.php?sid=14
    Have things come far since then?

  11. Nathan Hopson said,

    January 19, 2012 @ 8:06 am

    In response to Felix: yes (sort of). It's called back-translation (hyphen optional?).

    I still do this for Japanese translations of American / WHO medical protocols. I get a Japanese translation of the English original, and without consulting the original translate the Japanese manuscript into English. In addition to my comments about the quality of the Japanese per se, the original document creator compares his/her own English with mine to help determine whether the Japanese correctly translates the English. If potentially troublesome disparities are noticed, the translator and document author then confer, and in extreme cases the process is repeated.

  12. Ellen K. said,

    January 19, 2012 @ 8:20 am

    I use Google Translate to read and understand things written in Norwegian, and I think it does an impressively good job. Not good enough, though, for me to say "really good". "Pretty good", but not "really good". And it does better at Norwegian than at Chinese.

    I do wonder if that someone was using it for some language besides Chinese. And even if so, I think "really good" is optimistic.

  13. Neil Kubler said,

    January 19, 2012 @ 9:47 am

    Reminds me of the story told by one of my grad school mentors, the linguist Professor Charles Hockett, of the Biblical sentence "The spirit is willing but the flesh is weak," which was translated from English into multiple languages and then back into English, coming out as: "The ghost wanted to, but the meat was soft."

  14. Jean-Sébastien Girard said,

    January 19, 2012 @ 10:52 am

    @Felix FWIW, a mean of evaluation considered reliable and efficient across the scope of potential translations remains elusive. Backtranslation is relatively reliable, but it arguably fails miserably on the efficiency by requiring a translation to be done twice, which obviously the many freelance translators can't afford to do, nor can (typically) their clients.

  15. Dan T. said,

    January 19, 2012 @ 11:33 am

    The second sentence, which converges on "A heavy white mist, rising from below, to cover all of the vision.", actually stays pretty close to the original sense of there being a rising mist blocking your view, though it somehow becomes a verbless sentence fragment.

    The translation seems to have a problem with prepositions, with the first sentence mutating from you being on the side of the chasm to your actually being the chasm (or its side). Also, singulars and plurals keep flipping back and forth in successive translations.

  16. Ken Brown said,

    January 19, 2012 @ 11:39 am

    @Bears "That would be a most interesting game of "broken telephone!""

    Is Broken Telephone the same as what we call Chinese Whispers? ("Send three and fourpence, we're going to a dance")

    @Neil Kubler, I thought the canonical version was supposed to be via Russian and it came out as "The vodka was good but the meat was off"!

  17. Dan Lufkin said,

    January 19, 2012 @ 12:02 pm

    I use Google Translate in language pairs where there exist large corpuses of nearly parallel text. Here's a Danish –> English sample that just happened to be on my screen this morning:

    Grønlænderne har imidlertid to sprogbyrder: grønlandsk er en stor byrde og dertil kommer så dansk, som en betydlig sekundær byrde. Hvor mange procent af BNP denne dobbelte byrde koster Grønland, kan man kun gisne om, men det er ikke småpenge. Hvis vi siger, at byrden er 2% i Danmark, må den være over 5% i Grønland.

    Greenlanders have, however, two language burdens: Greenlandic is a great burden and then add to this Danish, as a significant secondary burden. How many percent of GNP this double burden costs Greenland, one can only guess, but it's not small change. If we say that the load is 2% in Denmark, it must be over 5% in Greenland.

    I touched up the translation very slightly, using the click-on feature of GT to get a few tweaks like "Greenlandic" instead of "Greenland" and "language" instead of "languages" as it read initially (Danish is chronically short of plurals). None of these tweaks took any particular knowledge and were actually prompted by GT. GT works great for Afrikaans English because practically everything that gets printed in AF also appears in EN. GT even gets the cockamamie AF negation structure.

    Where the Chinese Telephone gag usually breaks down (to judge by the blog samples) is in moving randomly between languages where there's practically no parallel corpus text. Basque –> Bulgarian is often the glitch. I guess the Telephoners are taking the available languages in alphabetical order.

    In short, I find that GT is a useful aid. If you don't know the target language, at least you get the gist and can decide whether or not it's worth paying for a professional translation. Just don't use it out of the box for something like a contract or patent or sales brochure.

  18. Jerry Friedman said,

    January 19, 2012 @ 12:38 pm

    @Ken Brown and Neil Kubler: My "canonical" version involves only Russian, and the result was "The ghost is willing, but the meat is raw."

    And "telephone" is the name I know for the game the British call "Chinese whispers".

  19. Ray Girvan said,

    January 19, 2012 @ 12:50 pm

    @Ken Brown and Neil Kubler and Jerry Friedman:
    My favourite in this genre is the alleged translation, from Scots to German and back to English, of Robert Burns's address to the haggis …

    Great chieftain o' the puddin-race!

    … that is said to have come back as …

    "Mighty Führer of the sausage people!"

  20. john said,

    January 19, 2012 @ 1:15 pm

    I'd like to see Victor Mair's translation of that text

  21. Ginger Yellow said,

    January 19, 2012 @ 1:21 pm

    Has the sadly defunct Translation Party really never been discussed here at LL?

  22. Nathan Myers said,

    January 19, 2012 @ 1:30 pm

    The end result really does resemble translations from the Tao Te Ching, and also what I have seen of the foundational documents of Chinese medicine. I am becoming convinced that people who insist they can read these old texts are fooling themselves, and that their interpretations amount to confabulations that construct a plausible story from more or less random input. (Add to such misunderstanding the likelihood of actual transcription errors.)

    Has this been expressed anywhere formal? I can understand that publishing on it might make it difficult to get access to old texts, just as does questioning the officially assigned dates of events in antiquity.

  23. David Eddyshaw said,

    January 19, 2012 @ 1:47 pm

    I first heard the English-Russian-English one as resulting in:

    "The vodka is OK but the meat has gone off."

  24. Victor Mair said,

    January 19, 2012 @ 1:49 pm

    @Nathan Myers

    Brilliant insight!!!

    Rudolf Wagner has written a couple of excellent papers demonstrating exactly what you suspect, namely, that many Sinological translations of ancient texts (both by Chinese and by Western scholars) amount to sheer gibberish because of profoundly mistaken notions about the nature of the original texts. One is in the Festschrift for Wolfgang Kubin and the other is in the Festschrift for Christoph Harbsmeier.

    If someone will promise to put these two papers up on the Web for everyone to see should they wish to do so, I'll send them right away (well, after I have lunch and take the train to the office!!).

  25. David Eddyshaw said,

    January 19, 2012 @ 1:52 pm

    @Ken Brown: Sorry, didn't read your post carefully enough …

    There's also of course the old chestnut

    "Out of sight, out of mind" -> "Invisible lunatic"

  26. LDavidH said,

    January 19, 2012 @ 2:52 pm

    @Brindi: Çkemi! (Or don't Albanians in the diaspora use "çkemi" as a greeting?) There are quite a few foreigners in Albania who learn to speak the language; very few (if any) manage to master the grammar. I think I'm probably the closest; I actually wrote an Albanian grammar for foreigners, and I wrote three books in Albanian (not to mention that I translated a number of worship songs). But I still make mistakes; as a matter of fact I still make mistakes in English, despite it being my second language and my wife's mother tongue!

  27. AntC said,

    January 19, 2012 @ 5:07 pm

    @Felix, @Bears, @others: [Paraphrasing from memory] Cervantes' prologue to Don Quixote claims that the whole story is based on a traveller's tale in Arabic. He got it translated into Spanish. And in case anybody should doubt the accuracy (not, of course, that any word of it rings untrue), he got the Spanish translated back to Arabic by a different translator. The back-translation matched word-for-word.
    I read this in the Penguin Classics English translation; since (sadly) my command of Old Castilian is not up to vouchsafing the 'original'.

  28. Jerry Friedman said,

    January 19, 2012 @ 5:43 pm

    I decided to take up Felix's challenge, since I have better things to do. But my performance shouldn't be taken as what a professional translator would do. I chose a children's book I haven't read that's out of copyright. Here's my attempt at re-translating the French translation by Varlet. I admit I knew the original of the last two lines.

    It is upon the urging of the knight Trelawney, Dr. Livesy, and all of those gentlemen in general that I have decided to put in writing all that I know concerning Treasure Island, from A to Z, omitting nothing but the position of the island, and that only because part of the treasure is still there. Thus I take up my pen in this year of grace 17–, and begin my tale at the period when my father ran Admiral Benbow's inn [?], on the day when the old sailor with the sun-burned and sabre-scarred face came to take lodgings under our roof.

    I recall it as if it were yesterday. He came with heavy tread to the door of the inn, followed by his trunk on a cart. He was a big sturdy fellow, with very brown hair twisted into a greasy braid that fell onto the collar of a filthy blue suit; he had hands seamed with scars, torn black fingernails, and the livid, dirty-white sabre scar spread across his cheek. Whistling all the while, he crossed the cove of our gaze [what?], then in his strident and quavering old voice, which the of the capstan had cadenced [?] and broken, he broke into the ancient chantey that he was to sing to us so often later:

    Fifteen men on a dead man's chest–
    Yo-ho-ho and bottle of rum!

    Here's the original:

    "Squire Trelawney, Dr. Livesey, and the rest of these gentlemen having asked me to write down the whole particulars about Treasure Island, from the beginning to the end, keeping nothing back but the bearings of the island, and that only because there is still treasure not yet lifted, I take up my pen in the year of grace 17—, and go back to the time when my father kept the 'Admiral Benbow' inn, and the brown old seaman, with the sabre cut, first took up his lodging under our roof.

    "I remember him as if it were yesterday, as he came plodding to the inn door, his sea-chest following behind him in a hand-barrow; a tall, strong, heavy, nut-brown man; his tarry pigtail falling over the shoulders of his soiled blue coat; his hands ragged and scarred, with black, broken nails; and the sabre cut across one cheek, a dirty, livid white. I remember him looking round the cove and whistling to himself as he did so, and then breaking out in that old sea-song that he sang so often afterwards:—

    "'Fifteen men on the dead man's chest—
    Yo-ho-ho, and a bottle of rum!'"

    Comments of all kinds are welcome. You'll notice I misunderstood a couple of things in the French and forgot "pigtail", a word I should have remembered, but I'm not taking the blame for everything. Is the description of the old sea-dog's voice in some other edition?

  29. KWillets said,

    January 19, 2012 @ 5:45 pm

    It might be interesting to write an article on why these translations fail; I've been reading through some notes by Och (http://www.mt-archive.info/MTS-2005-Och.pdf), and he confirms what I suspected: the algorithm is heavily dependent on n-gram matches in the training texts, for n = 1-4.

    I haven't had time to work it out fully, but my guess is that, because of word order, language pairs with very different grammars (eg SVO vs. SOV) tend to break down the length of the n-gram matches, so they become more like single-word substitution.

  30. Jerry Friedman said,

    January 19, 2012 @ 5:45 pm

    "…which the labours of the capstan…", maybe.

  31. Jerry Friedman said,

    January 19, 2012 @ 6:15 pm

    I see, I should have looked after the song for the description of his voice.

  32. David Eddyshaw said,

    January 19, 2012 @ 6:23 pm

    @KWillets:

    I've certainly noticed that the Japanese-English translations generally seem to be not much good, compared with English translations from common European languages. It's not hard to think of reasons why this could be so, but the SOV word order of Japanese had previously occurred to me as one possible contributing factor. If the Chinese-English translations are mostly better than the Japanese-English, that might well be part of the reason.

  33. Victor Mair said,

    January 19, 2012 @ 8:19 pm

    One of the many difficult aspects of learning Mandarin for speakers of English is to distinguish between shì 是 and zài 在, both of which are commonly translated as "am, is, are", whereas the former is a copulative verb and the latter is a locational verb. Google Translate's troubles with the passage selected began when it opted for shì 是 over zài 在 as the second word of the translation — a fundamental error that all good teachers of Mandarin strive to correct during the early stages of study.

    Another fatal error occurred when Google Translate rendered the second chasm as "information gap". Still, I found that to be so charmingly emblematic of the problems encountered in machine translation that I chose it for the title of this post.

    As an experiment, I asked three graduate students from the PRC to translate the original passage into Chinese, after which I had Google Translate render their translations back into English and I did the same. While I didn't repeat the process three times back and forth as in the original exercise, it is clear that the human translators stick much more closely than Google Translate to the meaning of the original passage, so that, even if I were to repeat the process (English –> Chinese –> English –> Chinese, etc.) numerous times, I doubt that the results would stray very much from the content of the original text (except when an outright mistake is made by the human translator, as with the very last word of the third translation).

    Gianni Wan:

    Nǐ zài yīgè dà shēn lièkǒu de yī cè. Yī tuán cóng xiàfāng shēng qǐ de nóngzhòng de bái wù zhēdǎngle quánbù tōng wǎng yuǎnfāng de shìxiàn. Yītiáo xīnán fāng de xiǎolù tōng xiàng yītiáo yuǎn lí lièkǒu de wānyán de zǒuláng. 你在一个大深裂口的一側。一團從下方升起的濃重的白霧遮擋了全部通往遠方的視線。一條西南方的小路通向一條遠離裂口的蜿蜒的走廊。

    Back to English (by Google Translate):

    You are in a large deep cracks in the side. A group from the bottom of a thick white mist rising block all access to the distant line of sight. A southwest path leading away from the tears of a winding corridor

    Back to English (by Victor Mair, done without reference to the original):

    You are on one side of a large, deep chasm. A dense, white mist rises up and completely blocks the line of vision into the distance. A small path to the southwest leads to a winding corridor that is far away from the chasm.

    Rebecca Fu:

    Nǐ zhèngshēn chǔ yīgè jùdà yōushēn de xiágǔ de yībiān. Báisè de nóngwù cóng xiàmiàn shēng qǐ, yuǎnfāng de shìyě móhú bù qīng. Yītiáo yánshēn xiàng xīnán fāng de xiǎo jìng, yǐndǎo nǐ líkāi xiágǔ, zǒu jìn yīgè wānyán de zǒuláng. 你正身處一個巨大幽深的峽谷的一邊。白色的濃霧從下面升起,遠方的視野模糊不清。一條延伸向西南方的小徑,引導你離開峽谷,走進一個蜿蜒的走廊。

    Back to English (by Google Translate):

    You are living in a huge deep canyon side. White fog rising from below the distance vision blurred. An extension of the south west path, and guide you to leave the canyon, into a winding corridor.

    Back to English (by Victor Mair, done without reference to the original):

    You are situated at one side of an enormous, deep canyon. A thick, white mist rises up from the bottom; your field of vision into the distance is blurred. A small path extending toward the southwest leads you out of the canyon, from which you enter a winding corridor.

    Jing Wen:

    Nǐ zài jùdà shēnyuān de yī cè, báisè nóng wù cóng shēnyuān zhōng shēng qǐ, shǐ duì'àn de yīqiè dōu móhú bù qīng. Yītiáo xiǎolù cóng shēnyuān xiàng xīnán fāngxiàng yánshēn guòqù, jìnrù yītiáo wānyán de suìdào. 你在巨大深渊的一侧,白色浓雾从深渊中升起,使对岸的一切都模糊不清。一条小路从深渊向西南方向延申过去,进入一条蜿蜒的隧道。

    Back to English (by Google Translate):

    In the vast abyss of your side of the white fog rises from the abyss, so that the other side of everything blurred. A path from the abyss southwest Fangxiang Yan Shen in the past, into a winding tunnel.

    Back to English (by Victor Mair, done without reference to the original):

    You are on one side of an enormous abyss. A thick, white mist rises up from the abyss, causing eveything on the opposite side to be blurred. A small road extends from the southwest of the abyss and enters a winding tunnel.

  34. Just another Peter said,

    January 19, 2012 @ 8:33 pm

    The (now finished) Australian TV show Spicks and Specks had a game called "Turning Japanese" in which song lyrics were translated into Japanese and back using an online translator and the contestants had to guess what the original song was.

  35. Victor Mair said,

    January 19, 2012 @ 8:46 pm

    @Ginger Yellow

    I recall that Translation Party has been mentioned in comments on earlier posts. However, unlike the present case, which deals with a paragraph-length passage, my impression is that Translation Party dealt mainly (only?) with short phrases or single sentences. Since the paragraph repeatedly manipulated in the present exercise is much more complex than the short phrases or sentences of Translation Party, the ways in which and the extent to which things can go awry are of greater extent.

  36. AntC said,

    January 19, 2012 @ 10:15 pm

    You don't even have to leave a single language to achieve some of those mix-ups. The fairy tale about Cinderella's glass slipper — think about it! how could you even walk in glass slippers, let alone dance, without ending up in casualty? — is thanks to confusing vair (finest kid leather) with verre (glass).
    (Apologies if my memory is unreliable: I can't find an on-line resource to verify the translation for vair.)
    I'd always assumed this was a bad translation, but no: it was a Frenchman. See http://fr.wikipedia.org/wiki/Controverse_sur_la_composition_des_pantoufles_de_Cendrillon

  37. Mat Bettinson said,

    January 19, 2012 @ 10:46 pm

    As a student of Chinese I've spent a lot of time with Google Translate. It's rather easier to get a wrong answer than it is a correct one.

    The locative 在 can obviously also be used to indicate the continuing action of the following verb. Google doesn't seem to get this, but it does get it when you use 正在 which is a more formal version. The task is hard, so making mistakes is one thing, but really 在 followed by a verb is not ambiguous.

    The more I've played with this, the more it's obvious that formal Chinese stands a much better chance than informal spoken-style Chinese. I imagine this is due to the training material and of course Google's pathological determination that no linguistic knowledge should be applied at all, even if for the purposes of sanity checking.

    One often makes fun of the prevalence of bad English in China as a result of the use of Google translate but interestingly I've also seen the reverse. My local tourist attraction has a translation of This way to … with the Chinese helpfully being something like 这种方式(Puffing Billy). Which is pretty funny, it's like saying: In this way Puffing Billy, also vaguely poetic I suppose :)

  38. Joe Fineman said,

    January 19, 2012 @ 11:22 pm

    Another probably apocryphal anecdote (supposedly a telegram translated from English to Russian to English): Your son suspended for minor offenses -> Your son hanged for juvenile crimes.

  39. Matt McIrvin said,

    January 20, 2012 @ 12:41 am

    In Philip K. Dick's 1969 science-fiction novel Galactic Pot-Healer, there are bored office workers who goof off over some kind of global communications network by playing a game in which one player translates and reverse-translates phrases using networked machine-translation servers, and another player tries to guess what the original phrase was.

    It's probably the most prescient passage ever to appear in a Dick novel (and the rest of the novel is typically wild Dickian craziness, so it really stands out).

    But I've always assumed he was riffing on those jokey anecdotes about early machine-translation efforts ("the vodka's good but the meat is rancid", "invisible idiot"). I know they're old enough that, whether or not they have any basis in fact, I was hearing them as old chestnuts about the cloddish literalism of computers back in the 1970s.

  40. Jerry Friedman said,

    January 20, 2012 @ 12:41 am

    @AntC: "Vair" is probably squirrel fur. The French Wikipedia article says the idea that the slippers were "vair" was first suggested by Balzac, long after Perrault wrote "verre". A Snopes piece points out that in a much earlier version of the legend, from China, the slipper is gold, which is also impractical. (This version is apparently "Ye Xian".) There are other versions (and other fairy tales) with diamond slippers, according to Cinderella: A Casebook. Eggcorns exist, but this doesn't seem to be one of them.

  41. Brindi said,

    January 20, 2012 @ 1:06 am

    @LDavidH
    Yes, they still use "çkemi" :) usually preceded by a "hë" when spoken aloud. It's wonderful you wrote a book on Albanian grammar for foreigners – there's not much of a selection out there. I make mistakes in English, too, often with idioms..sometimes I've never even heard them before. I realized this as I began having conversations with people older than me, who I guess were more prone to using idioms than friends my age were. Albanian is full of popular sayings and proverbs..I wish someone would collect those and put them in a book!

  42. AntC said,

    January 20, 2012 @ 2:50 am

    Thanks @Jerry F. I must be mis-remembering an apocryphal version. I'd vaguely inferred the squirrel fur from the French, but thought that equally implausible: prancing ponies; a crystal carriage; liveried footmen; a glittering ballgown; and … Squirrel fur? Moleskin? Rabbit pelts? Surely a fairy tale can find something more magical than that?
    The Snopes piece does, though, confirm my point that no mis-translation was needed.

  43. LDavidH said,

    January 20, 2012 @ 3:18 am

    @Brindi: I think they have. A friend of mine in Durres had a book with idioms and expressions. It was thick as a brick, and not very well organised, but still.

    On a translation confusion note: If I had written this in Albanian, you wouldn't have known if it was my friend or his book that was thick, since the pronoun would have been left unspoken!

  44. PaulB said,

    January 20, 2012 @ 5:49 am

    As P's Hungarian example shows, Google Translate performs poorly with names. One problem seems to be that it can't recognize declined forms of names unless they appear in its collection of parallel texts, which they usually won't. It needs to learn some rules.

    I talk about a Russian example here.

  45. Victor Mair said,

    January 21, 2012 @ 8:57 am

    Can anyone confirm or further identify this (from Bernhard Fuehrer)?

    ==========

    It reminds me of the poet and writer Guenter Eich (1907-1972). He had
    some knowledge of Chinese (I seem to remember he studied in Munich)
    and did a similar thing – hope I'm right in attributing that splendid
    and enjoyable piece of translation (from German to Chinese) and
    re-translation (from the Chinese back to German) to him.

  46. Victor Mair said,

    January 21, 2012 @ 9:32 am

    This one, form Sijie Ren, came in a bit late:

    Nǐ zài yīgè kuānguǎng de shēnyuān yī cè. Nóngzhòng de bái wù cóng dìdǐ zhēngténg ér qǐ, mísànle yuǎn chǔ de shìyě. Xīnán fāng yǒu yītiáo xiǎo jìng, tōng xiàng yuǎn lí shēnyuān de yītiáo wān yán láng dào. 你在一個寬廣的深淵一側。濃重的白霧從地底蒸騰而起,彌散了遠處的視野。西南方有一條小徑,通向遠離深淵的一條蜿蜒廊道。

    Back to English (by Google Translate):

    You are in a broad side of the abyss. Thick white mist into the sky from the ground transpiration, diffusion of the distant horizon. Southwest there is a path, leading away from the abyss of a winding corridor.

    Back to English (by Victor Mair, done without reference to the original):

    You are one side of a broad abyss. Thick, white mist rises up and spreads over the field of vision in the distance. In the southwest, a small path connects to a winding corridor that leaves the abyss in the distance.

  47. TSTS said,

    January 22, 2012 @ 4:28 pm

    Concerning Cindy Ning's use of Google Translate to look up vocabulary, I have also started doing that and feel it is quite useful. It makes mistakes but many of the mistakes are obvious enough to be caught by someone with basic knowledge of the other language.

    In fact, one major advantage of Google Translate over a dictionary is that you can provide context when necessary, by typing in a snippet of 3-4 words containing the term. And then maybe modifying that snippet a few times to learn when and how to use the word. I find this often more useful than the extra information offered by a conventional dictionary. Overall, Google Translate is a tremendous tool if used carefully.

  48. Keith said,

    January 23, 2012 @ 3:32 pm

    @AntC

    Vair is a way of combining the back fur and belly fur of a squirrel. The term is used in heraldry, and quite well explained in these two articles.

    http://fr.wikipedia.org/wiki/Fourrure_%28h%C3%A9raldique%29#Vair_et_Vair.C3.A9

    http://en.wikipedia.org/wiki/Vair

    As the French article points out, if your faery tale can have a pumpking turn into a carriage, and white mice become horses, why find it unbelievable that a girl dances in glass slippers?

    K.

  49. janet roberts said,

    January 26, 2012 @ 9:10 am

    a language poet would love this…lost in translation…in a mist clouds abyss canyons…makes me want to take a hike…through some gorge on a mountain; finding one's way into meaning, is about the same process, one step carefully in front of the other, eye on the way ahead, the intention, the attention, the meaning, the path, the goal….blue sky coming, translation, transliteration, literary transit. Thank you for all this; a great pleasure…a much better occupation…to not lose oneself in translation, or lose what is being said….of course, knowing both languages, the target and the original tongue well is key, isn't it…this could go on forever!

RSS feed for comments on this post