Become a fan of Slashdot on Facebook


Forgot your password?
IBM Communications Software

IBM Smartphone Software Translates 11 Languages 102

coondoggie writes to mention that IBM researchers have an internal smartphone software project that is capable of translating text between English and 11 other languages (Chinese, Korean, Japanese, French, Italian, Russian, German, Spanish, Portuguese, Italian and Arabic). There are no concrete plans to release this as a public product, but IBM certainly isn't shutting out that possibility. "Hosted as an internal IBM service since August 2008, n.Fluent offers a secure real-time translation tool that translates text in web pages, electronic documents, same-time instant message chats, and provides a BlackBerry mobile translation application. According to IBM, the software was developed from an internal IBM crowd-sourcing project where Big Blue's nearly 400,000 employees in more than 170 countries submit, update and continuously refine word translations. Every time it's used, n.Fluent 'learns' and improves its translation engine. To date, the tool has been used by IBMers to translate more than 40 million words, IBM stated."
This discussion has been archived. No new comments can be posted.

IBM Smartphone Software Translates 11 Languages

Comments Filter:
  • Wow! I'm afraid this is a very commodious.

    (forgive my terrible Japanese)

  • by Monkeedude1212 ( 1560403 ) on Monday November 23, 2009 @08:02PM (#30208410) Journal

    I can finally read that Japanese Slashdot?

    I've always wondered what crazy stuff goes on over there, I mean they are on the CUTTING EDGE.

  • by lamadude ( 1270542 ) on Monday November 23, 2009 @08:04PM (#30208430)
    the german phrase "Mein Luftkissenfahrzeug ist voller Aale" was correctly translated as: "My hoovercraft is full of eels" However the Hungarian translation was completely wrong
  • Dear Aunt...
  • Cool (Score:4, Funny)

    by pinkj ( 521155 ) on Monday November 23, 2009 @08:07PM (#30208464)
    J' hope for how that functions well!
  • If you give one of these phones to your girlfriend / wife, will it help you decode her rants into a language men can understand?

    • Re: (Score:3, Insightful)

      Hmm... "capable of translating text between English and 11 other languages (Chinese, Korean, Japanese, French, Italian, Russian, German, Spanish, Portuguese, Italian and Arabic)." Nope. Don't see Female on the list. Perhaps that's why they're not productizing it. Can it really be that useful if you can't understand (roughly) half the people on the planet?
    • Re: (Score:3, Funny)

      by baileydau ( 1037622 )

      Understand women. Not even God can do that... []

      It's an oldie ...

  • I'm a little confused about how this thing learns. A necessary component of learning is feedback, and I don't understand how this software will get any feedback correcting it when it makes some kind of translation mistake. Sure, the user could sit there correcting the output, but not only is that time-consuming, but also doesn't account for errors in translating TO the target language.

    I also suspect it must be some kind of cloud-based tool; one user's copy of n.Fluent improving itself wouldn't help anyon
  • 40 Million Words (Score:3, Insightful)

    by Nerdfest ( 867930 ) on Monday November 23, 2009 @08:32PM (#30208694)
    So basically, the fine print on one of their service contracts.
  • The ultimate test for machine translation system is whether it can do roundtrip translations without information loss or distortion of meaning. When I was in school somebody had carved "Borra mig i bjornen - Drill me in the bear - Drilla mig pa baret" in the desk. It's quite funny if you're a teenager and speak Swedish.
    • by isj ( 453011 )

      There will always be a risk of information loss or small distortion of meaning, because languages are not equal.

      Slang and sayings are probably the most difficult to translate. Eg. translating the English word "blue" to Italian will force you decide between "blu" and "azzurro" and either of those two choices insert extra meaning that wasn't in the original. Another example is the Danish saying "Træerne vokser ikke ind i himlen" which as far as I know has no direct equivalent in English.

    • "Hardness testing machine translation is whether you can do the translation, back and forth without loss or distortion of meaning. When I was in school, someone had written "Borras be with me - I am punching bear - bore me Baret bread" on the desktop. It is very nice if you are a teenager and speaks Swedish."

      That was a roundtrip through norwegian, catalan, german, swedish and estonian, with Google Translate. I think it's pretty impressive, only the carving has been significantly distorted - and that was arg
  • by sfcat ( 872532 ) on Monday November 23, 2009 @08:39PM (#30208762)
    What I want to know is if it can translate:

    The spirit is will but the flesh is weak.

    Other systems in the past has translated this English idiom into all sorts of laughable text but my favorite is

    The vodka is tempting, but the meat's a bit suspect

    There are many other famously wrong translations of idioms [] Admittedly, idioms are difficult to translate, but its not like the users will understand this or care. They just want a reasonable translation so they don't end up looking like an idiot to the cute foreign girl they are trying to bed.

    • Re: (Score:3, Insightful)

      by chill ( 34294 )

      Whenever anyone brings up machine translation there is always someone on Slashdot brings up this particular example, like it is some litmus test or something.

      I hate to say it, but I solved this one personally a few minutes after first seeing the problem. I noticed my computer had gigabytes of drive space, and I had a friend that was fluent in both Russian and English. I asked him to translate the phrase for me, the whipped up a perl script to give the correct translation.

      Considering computers are so good

      • Re: (Score:3, Insightful)

        by sfcat ( 872532 )
        Yes, you caught me. Some easy karma whoring but I still like the example and its good for a laugh. But idioms change with time and can be used in very fluid ways especially in social contexts. Basically what I'm saying is that computers still suck at understanding context with respect to natural language processing of all kinds. I've spent a considerable amount of my career trying to solve NLP problems but this one is a very tough nut to crack. But this doesn't change the fact that the user doesn't kno
        • by chill ( 34294 )


          But, why do computers have to solve it? Just damn well PREtranslate EVERYTHING and store it all in a lookup table. I mean, hire a crapload of people who are fluent in both languages and sit them down to translate newspapers, novels, speeches, and anything else they can get a hold of. Eventually, you'll have a database that covers the vast majority of all conversations in the target language. Anything it DOESN'T know it tries algorithmic translation and feeds to to a series of human translators for v

      • Yes, I'm sure any limitations of machine translation can be solved by an infinite series of corrective perl scripts.

      • Are you trying to be sarcastic? Gigabytes are not enough for any translation worth a penny. And Google translate is alive and well, it's definitively the best general machine translation system available anywhere (see my above post for a comparison to Babelfish).
        • by chill ( 34294 )

          What is wrong with gigabytes? Translations can be stored as compressed TEXT with . The whole text-to-speech and speech recognition are separate issues.

          And gigabytes was an example of cheapness of storage. Terabytes are a couple hundred $$ now. My point is storage is cheap.

    • by bronney ( 638318 ) on Monday November 23, 2009 @10:46PM (#30209630) Homepage

      Hey, looking like an idiot to the cute foreign girl is EXACTLY what's gonna get you in bed! :)

    • Re: (Score:1, Insightful)

      by Anonymous Coward

      How can you expect a machine to translate it right when a human can't get the spelling right?

      The spirit is willing but the flesh is weak.

      I think it applies to you perfectly, intelligence wise.

    • They just want a reasonable translation so they don't end up looking like an idiot to the cute foreign girl they are trying to bed.

      Guess what: So does the cute foreign girl! :P

    • by mcgrew ( 92797 ) *

      The spirit is willing but the flesh is weak.

      "Time flies when you're having fun". Why would I want to time flies? Especially if I'm having fun?

  • HIja' 'ach ta'taH 'oH ta' tlhIngan

    Machine translation courtesy of [].

  • If the software is calling a web service that performs the translation, then on the smartphone the software is trivial--a simple client that gets some user input, sends it to the internet, and receives translated text back. If this is the case, then there's no point in calling it "smartphone software", the brains are all on a server somewhere. And that server software deserves to be compared apples-to-apples to other online translation services like say... Babel Fish, to determine how worthy it is. Addin

  • Chinese, really? It translates things into a non-existent language? Maybe it can translate into Indian some day too.
    • by selven ( 1556643 )

      Chinese is an accepted blanket term for the dialects in that country. See []

    • They are most likely referring to Standard Chinese (also called Standard Mandarin), which is used in all government communications.
      • According to IBM, they are referring to " ... [conversion of] English to and from Arabic, simplified and traditional Chinese, French, German, Italian, Japanese, Korean, Portuguese, Russian and Spanish. ..."

        IBM also mentions that they have developed speech recognition sw for Hindi, one of the major languages of India. Speech Recognition, of course, is not translation, so it's not directly applicable to the parent post's topic.

    • by Frogg ( 27033 )

      now, i'm no expert in languages, but i do see that google translate [] also translate to/from chinese also, so i'm surprised that you claim it is a non-existant language?

      also, wikipedia have a page about the chinese language [] - whereas, conversely, and in support of the other half of your statement, they don't have a page for the indian language, instead having a page for the languages of india [].

      perhaps we differ over uses of semantics here? perhaps you would've been happier if they'd specified traditional or si

    • There is only one written language called "Chinese". Within that language, there are some simplified characters and traditional characters. Most mainland Chinese know mostly simplified forms (only of certain characters), and most other Chinese use the elaborate or traditional forms. This is basically due to the mainland Chinese effort to improve literacy by teaching primarily simplified characters when possible. However, they are just different forms for the words, and were used side-by-side historically in
  • i using! (Score:1, Funny)

    by Anonymous Coward

    Happy it is that I am to be informed of you that using translating device I slashdot egg war for screen.

    Last usable time once again for perfect reading!

  • If I read this properly this will be a web app, which will be nice for times when you have a data connection, but I have to wonder if you always will have that, the majority of the times when you need to use this.
  • by SashaMan ( 263632 ) on Monday November 23, 2009 @09:22PM (#30209086)

    My smartphone already does this - it's called google translate, and was a huge boon while I was overseas last month.

  • like this example of Hungarian [] to English translation.

  • by robwgibbons ( 1455507 ) on Monday November 23, 2009 @09:43PM (#30209236)
    When they couple it with spoken word recognition
  • i think ibm have some catching up to do! ;) - google translate [] does a lot more languages than that (51 in total) - in fact i'm kinda surprised google have not built it into their chromium-os or the android platform (erm, i dunno - maybe they have - it's difficult to keep up with it all)

    and, to top it all, google recently added [] the ability to view romanisations of characters such as chinese han, and input transliteration of phonetics for hindi, arabic and persian.

    to my technical yet non-linguistically educat

    • by mcgrew ( 92797 ) *

      although at least we don't speak americanese [/me ducks and runs]

      Actually there isn't an Americanese; there are a lot of dialects. There's TexMex, Ebonics, Redneck, etc. Coming form the midwest I understand someone from England far easier than someone from New England, whose dialect is nearly incomprehensible to me.

  • But an excellent translation of:

    "Finally made it to the middle class? So sorry, we are shutting down your shop and relocating your services to a less expensive country with even less paid drones. We're the new IBM... we don't make computers, operating systems. We just make it easier to manage slavery."

  • Its really useful software ibm put in mobile . its really help who want to learn other language and who travel around the world .awesome innovation for the users of mobile. Force Factor []
  • Accurate machine translation will never be achieved without the invention of human level AI. Translating from one language to another (Especially significantly different ones) requires full understanding of the contents of the text. It never could be and never will be achievable through word/phrase substitution.

    Language itself is full of ambiguities. Firstly, different languages have different ambiguities, choosing to encode different bits of information. Secondly, there are different usages of different
    • "And not all have a one to one mapping between the meanings of words in different languages. The same words in different languages can have very different connotations. Some words and phrases in different languages, you have no clear counterpart in other languages. These are often cultural specific conditions. A good translation must actively understand the meaning and purpose of the language. "

      That's Google translate, through English->Swedish->Chinese->Dutch->English.

      No human level AI, no under
      • by Jeeeb ( 1141117 )
        And not all about a sense of a word mapping between different languages. Same words in different languages can be very different connotations. There are words and phrases in different languages, you have no clear counterpart in other languages. These are often the culture of certain items. A good translation must be a positive meaning and purpose, how to use the expression.

        That's doing the same English -> Swedish -> Chinese -> Dutch -> English, as you did. For whatever reason my results are
  • I can't fathom that IBM wants to get into the Smart Phone business, being that they sold their ThinkPad business to Lenovo.

    However, selling this to Nokia, RIM, or whoever. Now that would make some sense.

    I would be a shame to see something like this die in their research labs.

    • I have seen some amazing, absolutely amazing things made by IBM and got wasted by "mainframe like" marketing.

      One of recent examples is XL Compiler stuff, last time I checked, some mainframe reseller was trying to sell it for $600 with horribly designed (front page!) page. Until PowerPC developers on Mac could trial it, damn Apple switched to Intel :) I use it as good example why that sad decision to switch to Intel was right thing.

      I have seen MPEG4 decoder/player written in Java, in JVM 1.1 ages. Imagine wh

  • It will weight a 800 pounds and have service where ever it bloody well pleases to have service.

  • I used the tool to translate a common Korean phrase that means "How's the weather today?" The n.Fluent English translation was "The fine weather today?" Since it allows you to "suggest" and submit a different translation, I did so. An hour later, I tried the same phrase, got the same response as before, and resubmitted the correct translation. An hour after that, I repeated the process, with the same results. Maybe it only lets managers make suggestions.

Neutrinos have bad breadth.