Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
Technology Science

Can Your Mouth Become Multilingual? 212

Roland Piquepaille writes "During a videoconference last week between Karlsruhe, Germany, and Carnegie Mellon University (CMU), Pittsburgh, USA, the talk of Alex Waibel, from CMU, was automatically translated in German and Spanish. Both the Pittsburgh Post-Gazette (PPG) and the Pittsburgh Tribune-Review (PTR) attended the conference, took pictures and were impressed by this new 'open domain' speech-to-speech translation. This new computer technology is based on artificial intelligence (AI) and statistical methods. During the demonstration, the speaker had electrodes attached to his face and his neck, but the researchers think that these electrodes could be implanted into your mouth and your throat in a decade from now -- if you agree of course."
This discussion has been archived. No new comments can be posted.

Can Your Mouth Become Multilingual?

Comments Filter:
  • Nifty but... (Score:5, Insightful)

    by Anonymous Coward on Sunday October 30, 2005 @05:48PM (#13911059)
    This still doesn't solve the problem that automatic translators still have problems processing the logic of certain languages. Just look at babelfish.
    • by Anonymous Coward
      ...processing the logic of certain languages

      Yeah, i'll be impressed when somebody makes software that can process the logic of perl. oh wait...
      • Re:Nifty but... (Score:2, Interesting)

        That's very different. It may seem complex, but the only reason your native tngue seems natural is that you have been exposed to it since you were born. Perl is very structured and strict as compared to a spoken language. I know it was a joke, but it is useful to note this point.
        • Re:Nifty but... (Score:3, Interesting)

          That's very different. It may seem complex, but the only reason your native tngue seems natural is that you have been exposed to it since you were born. Perl is very structured and strict as compared to a spoken language. I know it was a joke, but it is useful to note this point.

          Uh, my native tongue doesn't seem natural to me. Atleast, not anymore.

          I grew up speaking English, and at Tweleve I started learning Latin. Now I'm studying Japanese, and I've had extensive exposure to Spanish, Italian, and Germa

          • There are no human languages that are simple or unnatural. English is certainly cobbled together, but Japanese is mostly loanwords, too. The point is that there is no discernible difference in the difficulty of learning any human language.
    • by lintux ( 125434 ) <[slashdot] [at] [wilmer.gaast.net]> on Sunday October 30, 2005 @06:17PM (#13911230) Homepage
      Indeed. It still centuries last for computers really the capacity to approach sense construction of the human languages also but slightly of course.
    • This still does not solve the problem that automatic translators have still problems to process the logic of certain languages. Fair view of babelfish.
      • Wow. This is not what I need to read right now. Was just woken up out of a dream and my brain's still not fully awake. Seeing these grammatical monstrosities, I knew that A)something was wrong and that B)it made no sense to me. I couldn't figure out why this was, and was convinced by my dream-state pseudo logic that the reason this thread was making so little sense was that I wasn't fully awake yet. Eventually the real answer dawned on me that the thread was just people writing mangled english to make
    • I got my bachelor's degree in Linguistics at the University of Washington and one of the topics we worked on heavily was syntax structure and computational linguistics. The driving force behind much of that department is to progress the knowledge of language to the point where it can be completely digitized.

      The problem with Star-Trek-like speech converters is not in algorithms, language itself, or the computer models we use to represent it. The problem with perfect speech translation is language itself.
    • processing the logic of certain languages

      Yeah, like English. "I had my car fixed." - so, did you have it repaired, or made to not move? Determining the right definition from context is hard enough for people to do, let alone design a program to do it for us.

      And does the word "wind" refer to womething twisting, or the air movements? Consider "the wind winded along." Heaven help you to figure out what pronunciation, and therefore definiton, you were trying to get.

      'Natural language' is a really nasty set

  • Absolutely! (Score:5, Funny)

    by DoraLives ( 622001 ) on Sunday October 30, 2005 @05:49PM (#13911064)
    the researchers think that these electrodes could be implanted into your mouth and your throat in a decade from now

    Oh yeah. Please. Right now. Insert the electrodes into me right now. I can't wait!

    • by Anonymous Coward

            Make sure to throw in a couple of RFIDs in my anus while your at it.
    • by arose ( 644256 ) on Sunday October 30, 2005 @06:27PM (#13911289)
      Careful, sarcasm isn't translated by this system.
    • Oh yeah. Please. Right now. Insert the electrodes into me right now. I can't wait!

      Electrodes? how 1960.

      I don't understand why the passive approach of speech recognition is being overlooked.

      Speech recognition software has come leaps and bounds since the last decade. The only problem remaining is maintaining a constant during the process of feeding the sound into the processor; a microphone issue.

      I bet of you trained your speech recognizer with a microphone limiter would make all the difference.
  • by LeonGeeste ( 917243 ) * on Sunday October 30, 2005 @05:49PM (#13911067) Journal
    Don't believe a word of this. Everybody likes to say they've finally cracked the problem of machine translation. This is exactly what we saw previously on Slashdot with the quote about the "bin Laden tapes" or whatever.

    Proof? Ah, we'll get to that later.

    Where in any of the links does it give the text of what he said, and the translation? And the analysis to the success of the translation? I found two sentences it mentioned. That's not good enough. Let's allow independent examination of the success of this translation.
  • eh? (Score:2, Insightful)

    by clragon ( 923326 )
    constant maintence would be needed to maintain a good database for this translator, or else the 70s slangs would be coming out of the other guys mouth. it would be much easier to LEARN a language at a young age, then you dont have to worry about "are they understanding what im saying right now?" when you use one of those high-tech translators :D
    • Re:eh? (Score:2, Interesting)

      That happens with real people, too =)
      I know a Malay who learned English in Australia, and he talks like someone straight out of the 1800s.
    • OK, so you learn the language at a young age. Let's say Russia is a big concern at the time (Cold War is in full swing), and then you come to today where Russian is not quite as important. Instead, you need to know Arabic. So, kids today start learning Arabic. The problem is, by the time they grow up, Arabic isn't that usefull anymore. Instead, it's Mandarin. So, their children get to learning Mandarin, but by the time they grow up China is no longer the issue. Instead, they really need speakers of Tagalog
      • OK, so you learn the language at a young age. Let's say Russia is a big concern at the time (Cold War is in full swing), and then you come to today where Russian is not quite as important. Instead, you need to know Arabic. So, kids today start learning Arabic. The problem is, by the time they grow up, Arabic isn't that usefull anymore. Instead, it's Mandarin. So, their children get to learning Mandarin, but by the time they grow up China is no longer the issue. Instead, they really need speakers of Tagalog
        • The above poster had a good point. When I was growing up, Spanish wasn't that important. There was no compelling reason to learn it. 20 years later, most of the businesses have signs that say "Hablamos Espanol", and there is a bilingual neighborhood newspaper. Maybe in another 20 years we'll be a trilingual society with Mandarin or Arabic rivaling Spanish for the #2 language.
          • When I was growing up, Spanish wasn't that important. There was no compelling reason to learn it. 20 years later, most of the businesses have signs that say "Hablamos Espanol", and there is a bilingual neighborhood newspaper. Maybe in another 20 years we'll be a trilingual society with Mandarin or Arabic rivaling Spanish for the #2 language.

            Where did you grow up? In California, Arizona, New Mexico, and Texas, there have been significant Hispanic populations since before any English speakers were there.

            • Where did you grow up? In California, Arizona, New Mexico, and Texas, there have been significant Hispanic populations since before any English speakers were there. New Mexico is officially a bilingual state (bet you didn't know that). But I agree that the Hispanic population in the US has grown tremendously, recently, and even in the mid-west there are more and more signs in Spanish, more cable and radio stations in Spanish, etc. But as they said in West Side Story, Puerto Rico's in America, and it's been

              • t would be interesting if Portuguese ever became popular in the US. If I ever learn Spanish (I'm working on it, but it's going slowly), Portuguese should be easy, and if I knew those 3 languages, I could get by anywhere on 2 continents (except for Quebec).

                That's my rationale for learning Portuguese. Viva las Americas! (Besides Quebec, there are also Suriname [Dutch], French Guyana [French], and Paraguay [Guaraní] where people might not speak Spanish, English, or Portuguese, but that's nothing.)

  • by Quirk ( 36086 ) on Sunday October 30, 2005 @05:59PM (#13911135) Homepage Journal
    In terms of the hardware... " NASA scientists [astrobio.net] have begun to computerize human, silent reading using nerve signals in the throat that control speech." Subvocal speech reading systems offer the added bonus of now having to listen to the mundane trivia being broadcast from the cubicle next to yours.
  • by wertarbyte ( 811674 ) on Sunday October 30, 2005 @06:01PM (#13911147) Homepage
    So I won't even have to use a phone to get my speech intercepted, The Bad Guys [tm] now can directly wiretap my mouth. Perhaps this can be intercepted by chewing on my tinfoil hat?
  • by spongebue ( 925835 ) on Sunday October 30, 2005 @06:02PM (#13911155) Homepage
    Computers have a hard time translating written things as it is... any bilingual will tell you that online translators for complete sentences will do nobody any good, for the most part. My Spanish teachers are all able to see papers with computer translations very easily, due to similarities in words and meanings (such as the word "pants" which can be colthing or breathing heavily) Not to mention, grammar and things like that are not done well at all. For the fun of it, try going to an Online translator [freetranslation.com] and write something in English, translate it to Spanish, then back to English. Some results are pretty crazy. I guess the point I'm trying to make is this: what makes the translators so special compared to the ones we have now? How can they work better? Sure, there is probably a bit more effort put into these, but I don't think that a good translator will be available for another 5 years, not to mention the whole "take the speech you aren't saying" thing is hard to believe.
    • That's true, but their system was much more powerful. The first article made a quick mention of how "programmers try to make computers think like humans, while they were trying to make the computer work like a computer." I believe it said they were using statistics which (given enough source material) would eliminate those kind of problems. Seems like there was as article on Slashdot about that a while back.

      But by combining it with grammatical analysis you could also fix those kind of errors. In the exampl

      • Statistics can only help if you really have large corpuses of reference material, and feeding the text in two langage into the system will probably not be sufficient, you will need to map what expression goes to which. Gramatical analysis can only help to you a point. If you take this french sentence "Il va voler la vedette." It can mean either ''he will steal the show'' or ''he will steal the speedboat''. Statistics won't help you much: because one translation is more probable that the other doesn't make i
        • I live in Japan, too, and I completely agree that statistics is going to be only part of the solution. For example, take the following Japanese children's sentence:

          Inu wo mita.

          This could mean any of the following:

          I saw the dog.
          I watched the dog.
          I looked at the dog.
          I saw a dog.
          I watched a dog.
          I looked at a dog.
          I saw some dogs.
          I watched some dogs.
          I looked at some dogs.

          There are probably more as well. Some of these translations are more likely than others but they all depend on con
        • Finally, I must disagree about your comment about latin. The difficulty for X in speaking langage Y has a lot to do with how different they are in the logical structure, the vocabulary (and the sounds they use), so for people who speak romance languages, learning Latin will certainly be easier that say for a Chinese, people who speak langages with cases might also have an advantage

          I'd think Latin would be a total nightmare for a Chinese speaker. It's hard enough for English speakers, and we have a lot of L

      • by bodrell ( 665409 ) on Sunday October 30, 2005 @08:47PM (#13911978) Journal
        Of course, maybe we should all just switch to Latin. You can't say we're playing favorites with a language if you choose a dead one.

        Actually, that's exactly what Israel did when the Hebrew language was brought back from the dead. For awhile, German was considered for the official language of Israel, since there were so many German Jews relocated to Israel. A guy named Ben-Yehuda [wikipedia.org] was almost single-handedly responsible for reviving spoken Hebrew, making up Semitic-sounding words to fill in gaps, etc. Before that point, Hebrew was as dead as Latin (religious use only), although Yiddish [wikipedia.org] has a fair number of Hebrew words (and German, and Slavic).

        Besides German, I believe Russian and Yiddish were other popular choices for a national language, but each had its own political issues.

        No, I'm not Jewish. I just like languages.

        • Wow. Didn't know that. Very cool.

          I'd suggest something even less partisan like Esperanto (which has never been an official language anywhere) but since no one speak Esperanto as their only language, I know that it (like any designed language) would change when real people started to speak it. Latin has already been through that.

          Plus Latin used to be the "world language" back when all educated people spoke it an science was always done in it (witness Newton's "Principia Mathmatica", IIRC).

          But for the Hebr

      • I thought about this question and concluded that there were five levels of translation:

        1- word for word translation.

        2- phrase translation.

        3 - paragraph translation

        4 - conversational business level

        5 - diplmatic and literary level

        I suspect that each level requires an order of magnitude of more computing power than the previous level.

        By the scale above, I believe that Babelfish is on level two and SYSTRAN is on level three.

        I have used the SYSTRAN box to sell things on eBay to people in other countries that don
    • The computers have problems translating the things written as are... bilingual will tell him that translating in line for complete prayers they will do nobody good, for the the majority of the parts. My Spanish teachers are all capable to see roles with translations of computer very easily, due to similarities in words and meanings (just as the "pants" of word that can be colthing or they may be breathing a lot of) not to mention, the grammar and those things are not done well in all. For the amusing one
    • I just did exactly this (translate from English to German to English).

      I now understand how several of my coworkers write their emails.
    • Even something relatively simple...

      "This is a test of a translation from English to Spanish and back"

      Translated to Spanish: Esto es una prueba de una traducción de inglés a español y a la espalda.

      Which, back to English: This is a test of a translation of English to Spaniard and to the back.

      I'd really like to see them put their software up for testing. If it really is good enough to do speech in real time, it ought to be able to blow Babblefish and the like out of the water with on-

  • Smash Mouth (Score:2, Funny)

    by Tablizer ( 95088 )
    When I smash my thumb accidentally with a hammer, for a few seconds I can swear I am multilingual.
  • by G4from128k ( 686170 ) on Sunday October 30, 2005 @06:11PM (#13911199)
    Another approach is from some work I saw demoed at an MIT conference in Vienna. If you capture enough video of a person speaking, you can remix/rerender video of that person saying anything you want them to say [100777.com]. The software works at the phonetic level so you can even synthesize words that the person has never even uttered before and even make them appear to speak languages that they don't know. They had some visually convincing video showing people saying things that the researchers claimed they never said. Yes, the demo version worked with clean test video and a professional video/image analyst could probably spot a faked/remized video. But if these technology becomes good enough, I can see it making video a nontrustworthy source of data (like skillfully retouched photos).
    • They had some visually convincing video showing people saying things that the researchers claimed they never said.

      This is not exactly the same, but you should see this movie [tv-tokyo.co.jp] (get it with wget first, if it's not working).

      This technology is called 'Motion Portrait' and developed by Sony-Kihara Research Center that designed Graphics Synthesizer for PS2. It generates realtime (30fps) 3D facial animation from a single 2D picture, and is possible on an ordinary PC (you don't have to have a massive render far

  • by CyricZ ( 887944 ) on Sunday October 30, 2005 @06:17PM (#13911227)
    Indeed, we may see a new form of slander arise.

    Imagine what would happen if a malicious individual was able to modify such a system before a CEO gives a big speech to investors. The CEO is speaking English, but the Romanian and Chinese investors are listening in their native tongues.

    Soon enough the CEO is talking about synergistically-tiered multi-integrated doodads, but the Romanians are hearing "Cock sucking whore bitch! I fucked her up the ass in Bucharest and her nipples bled!", while the Chinese investors are hearing a whole string of racial epithets. Who would be responsible if such an incident occurred?

    Multiple nations also have hate crime legislation. Would the CEO now be responsible for committing a hate crime, merely because this device mistranslated what he said, and output racist remarks?

    • by JaredOfEuropa ( 526365 ) on Sunday October 30, 2005 @06:22PM (#13911257) Journal
      Would the CEO now be responsible for committing a hate crime, merely because this device mistranslated what he said, and output racist remarks?
      I will not repeat what Babelfish made of your post when it translated it into Dutch, but suffice to say you'll be hearing from my lawyer.
    • Obviously, no. IANAL, but most crimes are only crimes if you intentionally do whatever makes the crime (or try to, at least, but that already is something that must be specifically stated in the law, too).

      Just as an example, suppose you go to a fleamarket or yard sale or so and buy something from someone. If it turns out later on that that person was actually an imposter who took the money from you when the real owner wasn't there, does that make you a thief? Of course not. You'll probably have to give back
    • Shocking, I know, but the world has had both a) translators who have made mistakes or intentional errors and b) mechanical errors which greviously change the content of messages and we have somehow managed to muddle through. I'm not exactly an expert on the law in either Japan or the US (where I translate as part of my job, but its not my primary function), but my sense of it is that the translator has fiduciary duty to their client so you're almost certainly open to a tort if you intentionally try to scre
    • ... it is scratched [uibk.ac.at].
    • Are you serious?

      First, it isn't a crime in most nations to start cursing like a sailor.

      Second, even if it was a crime, it is pretty clear that if a translation program goes nuts and starts spouting obscenities, it is the translation program's fault. A tech might get sacked for failing to set up the program properly, but that is about it. That is like asking what would happen if someone hired a person to act as a translator and the translator started mistranslating things into obscenities. Uh, the transla
  • What about, I hack your mouth and say dirty things to your manager...

    or ..

    Update failed; you now got a speech impertinence. Don't mind the gaping mouth...

    or ..

    Download your speeches now; for 1500$ you can have your own personalized speech to your press conference...

    or ..

    This technology has already been invented secretly by the government and are testing it on the monkey in the BUSHes there...

    or ..

    Mass broadcast virus of speech; everyone starts saying sexist things to eachother...

    or ..

    The Jim Carrey or Robin
  • by ShahJehan ( 576346 ) on Sunday October 30, 2005 @06:37PM (#13911334)
    I am currently taking a great course on the Introduction of Linguistists. I have been exposed to the rather complex process a human being uses to make a sound (phoneme). You can go here to get a good idea of what it truly entails http://ocw.mit.edu/OcwWeb/Linguistics-and-Philosop hy/24-900Spring-2005/CourseHome/index.htm [mit.edu] The main obstacles to this is the fact each langauge uses different places and manners of articulation as well as the fact that intonation can change the meaning of a word. In Mandarin the word ma can change meaning based on tone. This is not a factor in English but certainly is for most Asian languages. The ability to use phonemes is one thing but paralinguistics is another (sarcasm).
    • As a Mandarin and English speaker in the US, I'm not sure I completely belive the part about intonation not affecting English. Take for example, my ... er... example:

      We're going to the store.
      We're going to the store?
      We're going to the store!

      In example one, it is common to give what would map to the "falling" tone in Mandarin on the last word. In the second example, it wouldn't be unheard of to use the "rising" tone. In the last example, one could use the "flat" tone to express excitement. In written Eng

    • We have two words in English which completely change their meaning depending on the tone of speaker. Those words are 'fuck' and 'shit'. The verb 'to fuck' basically means to have sex and the noun 'shit' literally means excrement.

      However the f word is used as a general purpose intensifier and the s word as a general noun for contempt. Or it can be any collection of undifferenciated objects.

      These words are illegal in most public mass media usages and are socially unacceptable
  • by 5i ( 112354 ) on Sunday October 30, 2005 @06:40PM (#13911357)
    If you're interested in reading the actual research paper involved (as opposed to a journalist's interpretation), it's readable here [cmu.edu] - pdf file, but lots of graphs, tables and pictures, so I'll forgive them.
  • no seriously. i was watching the extra stuff on "appleseed", and they talked about motion capture for expressions and talking. unfortunately, the motion capture actors had to keep their heads still for the scenes, necessitating the need to re-dub the movie due to the stiff acting. i know, it's not as earth shattering as world politics, but i thought it was cool.
  • Don't mean to lay bait, but like advanced sound and video editing, software for just text to text translation just ain't there (yes I searched apt-cache). Put a dent in my linux ego to have to bittorrent this sort warez for win xp.

    Take Debian -- Its users are spread out geographically about as even as something could get. You'd think that would be somehow conducive to developing this. What's the deal? I need something that handles more languages than google and babelfish.

  • by sbma44 ( 694130 ) on Sunday October 30, 2005 @07:16PM (#13911541)
    it's about speech recognition. They've identified a new source of data for identifying phonemes, one that apparently provides cleaner output than working from the audio. Dollars to donuts the resulting words are then just popped into a Babelfish-equivalent.

    This is interesting and important work, but the translation angle is really just one potential application of the technology.
    • You are correct, of course. Whether these muscle contractions are a more reliable source than the actual speech sound is not entirely clear. Also there must be huge variations between different people (just as in speech). The main problem, clearly, is that you have to fit people with electrodes and then tune the system. Seems a little far-fetched for a practical speech recognition solution as every time you re-fit, you would have to re-tune as well.
  • yes. (Score:4, Funny)

    by artifex2004 ( 766107 ) on Sunday October 30, 2005 @07:29PM (#13911593) Journal
    As soon as my girlfriend slips her tongue in my mouth, it becomes multi-tongued and French.

    And then I wake up.
  • by fraber ( 176289 ) <fraberNO@SPAMfraber.de> on Sunday October 30, 2005 @07:57PM (#13911761) Homepage Journal
    Hey,

    Alex Waibel was one of the leading scientists in the Verbmobil [verbmobil.dfki.de] project in 1995. The technology was pretty interesting (maintaining probability "graphs" from the Markov speech analysis through the syntactic and semantic analysis).

    However, results were pretty poor due to the structure of the project (just too many people) and because many institutions really weren't interested in the project and went for their favourite research topic with a new name (that's how research in Germany works...). Perfectly possible that Mr. Waibel advanced with the topic, now 10 years after the first major trial...

    Personally, I actually gave up AI completely after the ESSLLI (European Summer School on Logic, Language and Information) and promised not to touch the subject again until there were a "unified" formalism incorporating the old "symbolic" approach (predicate logic etc.) and the new statistical methods (Bayes, Markov, ...). Such a combination would be suitable both to deal with large amounts of data (statistical) and to deal with negation (only available in the symbolic appoach).

    Maybe they've got it this time? It's a pitty they don't talk more about the underlying formalism.

    Btw., the electrodes are probably just an enhancement of the normal speech recognition software to get a better "signal".

    Bests,
    Frank

    http://www.project-open.com/ [project-open.com]

    • Maybe they've got it this time?

      I'd just be happy if they got one step closer, and this work can be built upon. No one will ever 'get it', only because human language right now is too ambiguous. They'll get close enough to be useful though.

      However, my vision of the future is where human language and thinking actually adapts to the computers and becomes more formalized. The languages will be tightened up by standards committees and taught in schools as the official languages. After a few generations ther
  • "During the demonstration, the speaker had electrodes attached to his face and his neck, but the researchers think that these electrodes could be implanted into your mouth and your throat in a decade from now -- if you agree of course."

    *shakes head*

    NO NO NO, they've got it all wrong.. you implant a fish in your ear. That's how you speak multilingual, it's true.. i've read it in a book and even seen it in a movie (it must be true)
  • "the speaker had electrodes attached to his face and his neck, but the researchers think that these electrodes could be implanted into your mouth and your throat in a decade from now"

    Karl Rove of course ran the electrodes...

    Now we also know why Bush seemed so fried during one debate - the same electrodes give him a happy boost and also make him think God talks to him, so he can smite the Muslims.

  • Imagine someone hacking your personal translator while you are on a trip abroad:

    Hotel clerk: Rooms are 150 Euros a night.
    [Translated:] Rooms are 150 Euros a night.
    You: I won't pay over $130
    [Hack-Translated:] Deal
    Hotel clerk: Sign here

    I can think of many other scenarios, some funny, some sophomoric, and some downright evil.
  • by DroopyStonx ( 683090 ) on Sunday October 30, 2005 @09:50PM (#13912238)
    Why would you want to speak something other than American? Are you a terrorist?

  • We had this nifty thing at school years ago, enabled us to speak 4 different languages, it was called STUDY, in addition the more we understood all these foreign languages the better we came to understand our own.

After the last of 16 mounting screws has been removed from an access cover, it will be discovered that the wrong access cover has been removed.

Working...