back to article Microsoft demos real-time English to Chinese translation

In fiction Douglas Adams had his babel fish and Star Trek had the communicator, but Microsoft Research has been demoing an actual real-time English to Mandarin translation engine that works within seconds. Microsoft's chief research officer Rick Rashid showed off the technology at the company's Asian 21st Century Computing …


This topic is closed for new posts.
  1. Rusty 1


    When it can take "Hey, you sass that hoopy Ford Prefect? There's a frood who really knows where his towel is." and bounce it to Chinese and back without losing the subtleties of the original prose, I'll be impressed.

    Until then, I'm hanging on to my towel.

    1. LarsG

      Babel fish

      "Meanwhile, the poor Babel fish, by effectively removing all barriers to communication between different races and cultures, has caused more and bloodier wars than anything else in the history of creation."

    2. Anonymous Coward
      Anonymous Coward

      Re: Towels

      I'd like to see how it chews on this: -

      Oh freddled gruntbuggly thy micturations are to me

      As plurdled gabbleblotchits on a lurgid bee.

      Groop I implore thee my foonting turlingdromes

      And hooptiously drangle me with crinkly bindlewurdles,

      Or I will rend thee in the gobberwarts with my

      blurglecruncheon, see if I don't!

  2. Anonymous Coward
    Thumb Up

    This is where they make some good stuff

    Quite impressive.

    To be honest I've been quite impressed with some of Microsofts achievements on this terrain. For example; on my WP7.5 phone I have a (free) Microsoft app called "Translator".

    It allows you to translate text from stuff you typed using the keyboard, stuff from the camera (pictures where it'll detect text using OCR) or stuff you simply speak into the microphone.

    What I've been really impressed with was the other camera option: when I tell the app to translate from, say, English to Dutch after which I point the camera to my PC screen (when I'm typing this comment for example) then I can see the translated text appear on my screen in real time. It gets better: after the first run it will also detect some of the sentences it initially translated directly (one on one) while there are other sentence constructions for it. Those get changed over time too.

    As said I was very impressed with this app. Especially since it gets its information from the Internet (obviously but also allows you to download dictionaries so that you can perform translations without an Internet connection as well.

    Honestly; there are some fields where Microsoft can really be a trend setter. In my opinion of course.

  3. solidsoup

    So all the people who claimed we will all soon be learning Chinese to speak to our bosses can suck it. We'll be using automated translation.

  4. Anonymous Coward
    Anonymous Coward

    Translation from Market-speak to English:

    "We might have something that can compete with Google Translate's conversation mode. Maybe. We think."

  5. Charles Manning

    One word in 8?

    Sorry folks but you need a far better error rate than that before conference calls are going to happen without causing WW3.

    That 12% error rate was probably achieved with high quality audio. How well is it going to work with a crappy international line that drops the odd syllable now and then?

    There is no mention of the other way around. Is English easier to parse than Chinese? If so, it isn't much of a discussion.

    I expect to have my flying car way before this technology works reliably enough to put in the field.

    1. Anonymous Coward
      Anonymous Coward

      Re: One word in 8?

      Being a language where any particular word or phrase means only what the utterer indented it to mean and nothing more, it is impossible for a machine to parse it accurately. Unless that is it can also pick up on the thought processes and the state of mind of the originator. cf. any comment on The Register forums by amanfrommars1. Being human and having a reasonably broad knowledge of how the nuances of a spoken idea can be expressed in English text, I can make a stab in dark at what he means. For anything non-sentient it would be hopeless.

      I would assume by necessity, parsing Chinese is somewhat easier.

      1. Esskay

        Clippy Reborn

        HI THERE!

        It sounds like you're trying to abuse someone!

        Would you like me to:

        -insult their mother

        -insult their country

        -mention the war

      2. Allan George Dyer Silver badge

        Re: One word in 8?

        Parsing might be easier, but recognising the words would be a different problem. Chinese is a tonal language, Mandarin uses four tones, Cantonese more. The same sound said in a different tone is a different word. There's a Mandarin tounge-twister, "Shea shea shea shea shea shea shea", said with the right tones, it means "Forty-four stone lions".

        1. Arctic fox

          @Allan George Dyer RE " means "Forty-four stone lions"."

          I am old enough to remember that in the days before breathalysers were invented one quick way that some police used as a rapid judgement of a driver's condition was to get him to repeat "The Leith police dismisseth us" and if he couldn't then it was down to the station and the attendant police surgeon for a rather more scientific investigation. However, I have a gut feeling that the example you have posted is probably even more challenging stone cold sober, never mind a few sheets to the wind!

    2. Arctic fox

      Re: "One word in 8?" Yes? This kind of challenge is huge and even if this kit is still not......

      ..............ready for showtime it is still an impressive advance (as pointed out in the article). The technical difficulties involved in this kind of technology are enormous and this is a clear step forward. Not good enough of course but nonetheless a clear improvement.

  6. Esskay

    Different approaches

    I was always curious as to why a system similar to SwiftKey couldn't be used for voice - ie. looking at words grouped together as a sentence to "guess" what the last one probably was. In many ways, it seems it would be easier to improve accuracy with whole sentences than it would be when giving an AI simple commands like "open", "close", "delete", etc since there's context in a sentence that can be used to help "fill in the gaps" so to speak.

    Since there's still a delay of a few seconds in Microsoft's system, having a delay to look at words in a sentence before providing a translation wouldn't be too inconvenient. Also using a library built from previous calls (although it creates a privacy issue) or requiring the user to read out loud a few pages of a book (tedious but helpful) could help.

    At the very least it'll make the experience of getting a cab to your hotel in a foreign country after stepping off a 24 hour flight a bit less mental

  7. Frumious Bandersnatch Silver badge

    the time has come to talk ... of hovercrafts and eels?

    Sure, the MP sketch is a nice link to include, but wouldn't an omnigot link be more relevant to the topic? Still, I guess you have one now.

    1. Ian Michael Gumby Silver badge
      Thumb Up

      Re: the time has come to talk ... of hovercrafts and eels?

      ma dalag̃u gubi sugam

      It looks much cooler in Ancient Summarian.

  8. Frumious Bandersnatch Silver badge

    "Personally I believe this will lead to a better world."

    At the precise moment, the phrase "Personally I believe this will lead to a better world." (muttered by Rick Rashid to the assembled delegates, which for some strange reason was carried by a freak wormhole in space back in time to the farthest regions of the universe where the G'Gugvuntts and the Vl'hurgs lived) filled the air over the conference table, which in the Vl'hurg tongue was the most dreadful insult imaginable. It left them no choice but to declare war on the G'Gugvuntts, which went on for a few thousand years and decimated their entire galaxy.

  9. Andrew Garrard

    Digital babelfish, how I miss you

    Some remembering the earlier days of the internet may recall one of the first popular public digital translation services,, since moved to, and now redirecting to Bing translate. Altavista, of course, started out as, set up by Digital/DEC, and some of us remember when it was the search engine of choice.

    So, no - digital babel fish is not closer. Microsoft have moved Babelfish several step away from Digital.

    That said, I'm often astonished that Google Translate manages to produce something that's not a coherent sentence, let alone incorrect. I always assumed that something in the implementation of these things must understand some rules of grammar which ought to make that kind of problem tricky.

    1. M Gale

      Re: Digital babelfish, how I miss you

      Your problem is that a language is more than just words, and words is all a computer sees. Sometimes there is no direct translation and one word becomes twisted into a partial sentence in an attempt to convey the same meaning. See the German "schadenfruede", which translated to English is.. errr.. "schadenfruede".

      Then you add homonyms, synonyms and odd dialects of the same language, and what you have is a recipe for a programmer's nightmare. Multiply that by the number of languages in the world, and you can see the scale of the problem.

      Or just pump the phrase of your choice into here and have fun with the problem instead. Yes, yes I was looking for an excuse to post that link.

      1. hamsterjam

        Re: Digital babelfish, how I miss you

        An interesting fact: there are no English words for "Schadenfreude" or "Blitzkrieg". And there are no German words for "fair play" or "gentlemen's agreement".

        Make of that what you will...

  10. Best Before:

    Translation party

    haha I broke it with the phrase: "so long and thanks for all the fish."

    Coat duly collected.....

  11. Robert Helpmann?? Silver badge

    Star Trek had the communicator

    Yes, but it was the Universal Translator that did the work.

    What will make this a system worth using or one that will stagnate over time is how much its users can customize it and whether those customizations will be fed back to the main database (assuming there is such a thing) to improve the product.

    I am very familiar with the way text-based translation works and how OCR software can feed into it. Doing the same thing with the spoken word is at least a magnitude of difficulty greater. I do, however, wonder if it will "learn" to translate all the nasty bits of languages it handles first in much the same way most school kids do when confronted with learning a new language.

This topic is closed for new posts.

Biting the hand that feeds IT © 1998–2019