The wonders of Google Translate

« previous post | next post »

I have sung the praises of Google Translate (GT) before (e.g., "Google Translate is even better now" [9/27/16]), but this morning something happened with GT that really tickled my fancy.

One thing I use GT for is to compose texts in Chinese.  I find it to be a very powerful and easy to use input tool.

So I input the following:

shuō dìngle 說定了
xīngqítiān zhōngwǔ jiàn 星期天中午見

After I finished typing that, I glanced over to the box at the right where the automatic English translation appears.  I was just floored when I saw this:

That's a deal
See you at noon on Sunday

The GT translation is both idiomatic and natural.  Miraculously, it somehow even managed to catch the playful tone of what I wrote in Chinese.  Of course, when used irresponsibly by people who know no Chinese to check it or who try to get it to translate something that is literary / classical / topolectal when it is designed for Mandarin, it can produce Chinglish howlers.  But in this case (and in many other cases that I have experienced), GT is every bit as good as a human translator, and sometimes better than most.



17 Comments

  1. Colin McLarty said,

    September 22, 2017 @ 9:06 am

    I tell students constantly that, while they obviously should not use any tool stupidly, life is too short to *not* use Google Translate.

  2. Lupus753 said,

    September 22, 2017 @ 9:09 am

    Unfortunately, my Google Translate app seems to think that the text is in Vietnamese. I have no idea why. Do the Vietnamese even still use hanzi?

  3. unekdoud said,

    September 22, 2017 @ 9:25 am

    Two more wonders of Google Translate's Chinese (and Japanese) input: You can turn off its IME interface and just type pinyin without tone marks, and GT will give you its best guess and translation, and if you use the handwriting input method then after the first character GT will show you an autocomplete dropdown with translations just like the one it has for English.

  4. Mark Metcalf said,

    September 22, 2017 @ 11:53 am

    I usually check both translate.google.com and fanyi.baidu.com.
    The latter translated "說定了. 星期天中午見" as "Sure. See you at noon on Sunday"

  5. David Morris said,

    September 22, 2017 @ 6:53 pm

    My students were writing about a famous person they admire, and one started 'Jin Xing is a famous Chinese hoofer' – he'd simply copied the word from a dictionary/translator. I can't think that any standard dictionary/translator would give 'hoofer' as a choice *at all*, let alone as the first or only. This is assuming that he entered a standard Chinese word for 'dancer' and not one as slangy as 'hoofer' is in English.

  6. Colin McLarty said,

    September 22, 2017 @ 7:03 pm

    @ Mark Metcalf This agrees with my general experience that fanyi.baidu.com (going into English) gives noticeably more colloquial, natural results than translate.google.com; but is sometimes freer — less specifically close to the original.

  7. Victor Mair said,

    September 22, 2017 @ 9:36 pm

    "…my general experience that fanyi.baidu.com (going into English) gives noticeably more colloquial, natural results than translate.google.com"

    That has not been my experience at all. Baidu Fanyi (BF) is not bad, but for consistency of accuracy and idiomaticity, it can't come close to GT. If I only have time to check one online machine translator, I will always check GT first. BF is only for a backup.

  8. maidhc said,

    September 23, 2017 @ 3:16 am

    I had a very strange result from Google Translate a little while back. I had some Mandarin words (pinyin) but there were no tone markers (it was on a sign). I thought I'd put it in anyway to see what I'd get. GT just started spewing out reams of meaningless garbage.

  9. Victor Mair said,

    September 23, 2017 @ 6:33 am

    Not strange at all! This is a typical case of garbage in, garbage out. We had countless examples of it during a spate of LLog posts on computational linguistics "poetry" within the last half year or so (e.g., "Your gigantic crocodile!" and "Elephant semifics". Can't blame GT for that!

    http://languagelog.ldc.upenn.edu/nll/?p=32430

    http://languagelog.ldc.upenn.edu/nll/?p=33613

    Even with garbage, GT heroically tries its best, and sometimes comes up with "interesting" results.

  10. Rodger C said,

    September 23, 2017 @ 10:32 am

    Better a hoofer than a dotard.

  11. Reza.M said,

    September 23, 2017 @ 2:24 pm

    Google Translate is useful for get-aways where you'll just every so often require an interpreter and circumstances requiring just a straightforward interpretation. (More costly applications exist for basic, and more consistent, interpretation needs.) It's additionally amusing to play around with and a decent dialect device for kids.

  12. Tom Davidson said,

    September 23, 2017 @ 11:37 pm

    See what I found today on Google Translate:
    China's poor population of more than 8000 million people
    中文:我国农村贫困居民达8000多万人
    Source :https://translate.google.com/translate?hl=en&sl=zh-CN&u=http://blog.people.com.cn/article/1482981149339.html&prev=search

    One hundred million 40 million
    More than one million more than 4000 million
    中文: 一亿4000多万
    Source: https://translate.google.com/?safe=active&um=1&ie=UTF-8&hl=en&client=tw-ob#auto/en/%E4%B8%80%E4%BA%BF4000%E5%A4%9A%E4%B8%87

  13. Jonathan Smith said,

    September 24, 2017 @ 12:16 am

    @Tom Davidson
    and yet Google Translate
    "一亿4000万" => "140 million"
    which seemed encouraging (多 was the problem)
    only I removed the accidentally-included quotation marks and
    一亿4000万 => one hundred and forty thousand
    so back to "what the hell?"

  14. flow said,

    September 24, 2017 @ 8:35 am

    I don't think it's the 多 that throws GT off, it's the mixed-scripts numbers. When I put in 我国农村贫困居民达八千多万人 I get "China's rural poor population of more than 80 million people", and
    一亿四千多万 results in "More than 140 million", with correctly translated numbers in both cases.

  15. Tom Davidson said,

    September 24, 2017 @ 3:25 pm

    Then later in the day, looking up catch phrases for the 19th Party Congress, I found this: 劝返一批外逃人员”。
    (G) to persuade a group of foreigners.

  16. Daniel Lim said,

    September 27, 2017 @ 5:17 pm

    Thanks for pointing out the usefulness of Google Translate as an input tool for Chinese. I was not aware of this previously. Indeed it is easier to write Chinese with it than using the default input tools on Mac or Windows.

    To have some fun, I tried:
    说定了,明天我们去逛街。想了就乐。

    Google Translate gives:
    Definitely, tomorrow we go shopping. Think of the music.

    and Baidu Translate:
    Sure, we go shopping tomorrow. Happy is he who thinks

    Changing the last punctuation in the Chinese sentence to the exclamation mark, i.e.:
    说定了,明天我们去逛街。想了就乐!

    Google Translate gives:
    Definitely, tomorrow we go shopping. Think about it!

    and Baidu Translate:
    Sure, we go shopping tomorrow. When you think about it, you will be happy!

  17. PB said,

    October 1, 2017 @ 11:44 am

    Sometimes, Google Translate is still less than wonderful. I'm currently trying to research whether a small political party that participated in Iceland's elections in 2013, called Lýðræðisvaktin, still exists in some form (it's in any case certainly not very visibly active now). "Lýðræðisvaktin" is usually translated into English as "Democracy Watch". Well, this is an excerpt from a 2014 article in an Icelandic newspaper, talking about public funding received by the party:

    "Auk Bjartar framtíðar hafa Lýðræðisvaktin og Hreyfingin, sem nú heyrir sögunni til, sent Ríkisendurskoðun uppgjör vegna síðasta árs. Lýðræðisvaktin, sem ekki náði manni inn á þing fékk tvo styrki frá lögaðilum, upp á samtals 800 þúsund krónur. Ríflega 400 þúsund króna afgangur var hjá Lýðræðisvaktinni á síðasta ári."

    Google Translate translates this into English as follows:

    "In addition to the bright future, the democratic watchdog and the Movement, which now hears the story, have sent the National Audit Office a report last year. The democratic scourge, which did not reach a parliament, received two grants from legal entities totaling 800 thousand krónur. Over 400,000 krónur left at the Democratic Guard last year."

    So – the short text mentions the name of the party, Lýðræðisvaktin, three times. And each time Google Translate gives a different translation: "democratic watchdog", "democratic scourge", "democratic guard"…

    Well, I assume it's no wonder that it's still much harder for Google Translate to produce an accurate translation for a text in a small language such as Icelandic, dealing with a rather obscure political topic, than for everyday phrases such as "See you at noon on Sunday" in Chinese.

RSS feed for comments on this post