Follow Slashdot blog updates by subscribing to our blog RSS feed

 



Forgot your password?
typodupeerror
Technology

Wearable Translator to Debut at Comdex 93

quiller writes "Via is supposed to have a wearable PC that will take your voice, translate to seven different languages, and output the translated words through a speaker. Looks like something I want to look at while I'm there. " It will allegedly be showing at Lernout & Hauspie's booth, as it uses their translaton engine. The current specs have Mandarin Chinese, English, French, German, Italian, Portuguese, and Spanish in the box. I think I'd need one - I'd feel so Arthur Dent.
This discussion has been archived. No new comments can be posted.

Wearable Translator to Debut at Comdex

Comments Filter:
  • The Wearable PC is a line of products allowing the user to select processor, storage, and display depending on the application, and starts at $2,000 for quantity purchases.

    Storage and display choices. Is that really necessary? A simple LCD and a couple of RAM cards should be enough. But, what I want to know: is this thing real-time, or is there some (significant) delay in translation time?

    Universal Translator, meet the 20th Century.
  • Was it star trek or another show that had a "wearable transelator"?

    Does anyone know if there is a site that keeps track of sci-fi items that have become reality?
  • Imagine life actually being like a bad japanese movie with voiceovers!

    No one's voice syncing with their mouths would be very disconcerting. Too bad it doesn't do like the trek Universal Translator and change the apperance of their mouths to match.
  • Does anyone know what kind of accuracy and speed these things(or the engine behind the translation) are going to have?

    How extensive are the dictionaries?

    I haven't seen many impressive systems to translate text, let alone voice, so I am a little skeptical...
  • This is so cool. Imagine being able to mangle everything said to you by any foreigner, live and direct. No hunting through the dictionary, no browsing a web site just straightforward ...

    "I was pissing by the door and I heard a strange nuzze".

    ( Alo Alo for the humor impaird )

  • by Enoch Root ( 57473 ) on Thursday November 11, 1999 @09:55AM (#1541608)
    If this thing is as good as the Babelfish, I can just see it...

    Man walking down a street of Beijing: "I like the flawless beauty of the streets!"

    Translator: The virginity of these pretty [women] turns me on in the road!"

    "The wages of sin is death but so is the salary of virtue, and at least the evil get to go home early on Fridays."

  • "Why is my monitor blue?" into Mandarin Chinese...

    (For those who don't know or forgot, follow this Slashdot link. [slashdot.org]

    -m

  • how far back do you want to go? Jules Verne (submarines and flight), Leonardo DaVinci (tanks and helicopters) or even Nostradamus (sp?)

    I think this would be a great idea, but only if weighed in at the same time with all their ideas that never came about... sort of a shopping list. :)

  • PLEASE try the following words, to see what happens.... :) I'd -love- to know!

    • Ghoti (the linguists' favourite! :)
    • Floccinauccinilihilipilification (buffer overflow! :)
    • Antidisestablishmentarianism (Double negative. Syntax error at line 10. :)
    • Microsoft (Well, you just -have- to know what their name -really- means... :)
    • Linux (How does -this- gadget think it's pronounced? :)
  • Stick it in your ear.

    Now, how do you suppose it would translate that phrase? It is totally appropriate and polite when referring to the appearance of universal translators.

  • "...and I want to massage your grandmother."
    • -Steve Martin, commenting on speaking French badly
    If I used one of these, I'd want to know enough of the destination language to "sanity check" its output.
  • by jabber ( 13196 ) on Thursday November 11, 1999 @10:04AM (#1541614) Homepage
    The things that people [ufl.edu] have said..

    Ummm, anyone bite the wax tadpole lately?
  • Pricy, clumsy and arguably unnecessary. Choose a common second language - Esperanto, Ido, Cook Island Maori, who cares which one just so long as its not as twisted as English and make it the official SECOND language. Too bad so many folks are so attached to our "mother tongue" to be open to such a simple, elegant and efficient idea.

    Note: I said second language - I'm not advocating replacing anyones precious national language.

  • This sounds really neat, but the problems are many...
    • price
    • speech recognition - even pure English readers require the speaker to be clear and/or slow, and they aren't even always correct, and forget about it if there is a lot of background noise
    • translation - Babelfish [altavista.com] isn't the best, but it demonstrates that translation software still has a way to go before it replaces human translators

    However, if I could afford one of these, I suppose it would be better and faster than getting a pocket translator, typing in a phrase, and trying to pronounce the output.

  • ... in the honesty of the developers of the translation software. Does anyone remember Monty Python's "Bulgarian to English Phrasebook" sketch?

    "In this phrasebook, you have the Bulgarian expression 'Which way to the train station' translated as 'Please fondle my buttocks'...."

  • MarketingSpeak into plain English. It is Comdex after all.

    On second thought, maybe you just turn it off the get the translation.
  • by Joseph Vigneau ( 514 ) on Thursday November 11, 1999 @10:08AM (#1541619)
    The true test will be whether or not the thing can
    1. Listen to a human
    2. Output to another language.
    3. Have another device listen to that output,
    4. Output back to the original language.
    (In other words, box -i english -o mandarin < english.au | box -i mandarin -o english > english2.au) Only then will I be convinced that this isn't crap :^)

    -joev

  • by SmileyBen ( 56580 ) on Thursday November 11, 1999 @10:08AM (#1541620) Homepage
    ...you're not going to get the truly inter-planetary hitch-hikers wearing this thing. Fine it may do the job, but does it have the style, the finesse, of sticking a fish in your ear?

    I think not!!!
  • Knowing the difficulties of language translation and the current level of machine translation, there are just to many problems to make this feasable. The proof of concept in this device's creation is a strong example of the power of modern computing, but I feal that long term evaluation will show its considerable weakness. Cool gadget, but not practical. Spend your $2000 on a good translator.
  • " how far back do you want to go?"
    You could go as far back as the Bible, if you wanted. In the book of Acts, on the Day of Pentecost, St. Peter et al were speaking in their own language and their words were perfectly understandable to "all in their own tongue". But that one's usually chalked up to "Cool Miracle", not "Cool Technology" status.
  • by kevin805 ( 84623 ) on Thursday November 11, 1999 @10:11AM (#1541623) Homepage
    Machine translation isn't there yet. Babelfish at least starts with what you want in the first language. Imagine feeding the output from a 95% accurate speech recognition system though babelfish. You'd come out with gibberish.

    I'd prefer a system that assists me with speaking a second language. Something where I can be talking to someone, forget a word, hit a button on the Language eCoach(tm), say "you're welcome, in japanese" and hear "dou itashimashite" in a earphone. Or, someone says something I don't understand, so I repeat it to the translator and it gives me the english.

    With such a device, it would be possible to have a conversation in a foreign language after about 80 hours of instruction, because you don't have to memorize heavy vocabulary. It would also make the learning itself easier, because you don't have to waste time looking stuff up in the dictionary.

    If the translation was imperfect, it wouldn't matter so much. Maybe it'd give me different options, like if I say "bank, in german", it would say "with money, Bank, with river, Strand", and give the user the option of saying whatever is right. Babelfish translates "I went down to the bank" as "Ich ging unten zur Bank", which may or may not be what I meant.
  • "My hovercraft is full of eels."
  • Their website can be found here [via-pc.com].
  • Let us hoping to be and that this one is can work much better at translation than for babelfish.altavista.com. Aha, I wager you are looking to next, and yet, for success!!!!
    ------------------------------------- ---------------------
  • My hovercraft is full of eels!
  • Notice Japanese is suspiciously absent from the list of supported languages, but Mandarin Chinese is first. A Taiwanese product, no doubt, and perhaps if a certain neighbour to our east (north) might have raped our land 50 years ago we might be neglecting their language in our technology too. Still, let's not forget what happened to the collaborators who allowed the Japanese possession of Manchuria. . .

    Something to think about.
  • I like the idea of esperanto in a translater but why not use the language for its intended meaning. I'm trying to say, rather than one person having a
    English -to- Esperanto -to- french
    and the other person speak french,
    have one use a
    English -to- Esperanto (or visa-vis)
    and the other a
    French -to- Esperanto (or visa-vis).
    that would seem to simplify things. or am i just out of my knowladge league?
  • You can get the effect you suggest, can't you? Just turn the thing off until you need assistance. From what this thing is purported to do, it can do everything you suggest, since what you suggest is a subset of what it can do.

    Accuracy is another matter though. Even so I think it would be a cool thing for someone trying to learn a language.
  • by razvedchik ( 107358 ) on Thursday November 11, 1999 @10:58AM (#1541637)
    As a graduate of the arduous Russian Basic Course at the Defense Language Institute and one of the best American-born Russian speakers, I think that the whole of electronic translation is shallow and no substitute for going out and learning a language.

    When I've translated or interpreted (translation=written documents, interpreting=spoken in real-time), most of the time both of the parties have only a 75% clue as to what actually happened. They miss out on the connotation of the words, the hidden meanings that are derived from culture. In these cases, only the translator knows 100% of the transaction.

    For example, "Perestrojka" is the restructuring of the Soviet government during the Gorbachev era, but it comes from the roots "pere" or repeating action, again, and "strojit" or building, erecting, organizing. It's not just a political process, it's also what happens after an earthquake, and what I would call the Post-Civil-War Reconstruction if I had to talk about it in Russian. So, to an American, it is the policy of restructuring the government, but to me, it means a broad revolution of culture, ideas, and politics.

    Point being, that would be lost in an electronic translator. There are many concepts that don't translate no matter how hard you crunch code. You have to feel them.

    As Americans, we have this belief that everybody should learn English to talk to us. There's a joke in Linguist circles, "what do you call a person who can speak two languages? Bilingual. What do you call a person that can speak three languages? Trilingual. What do you call a person that can speak one language? American."

    Europeans have great language programs for school children, and it is no big deal for someone to learn Italian just because they are going on a vacation to Italy.

    If you want to go to a different country and buy souvenirs, get an electronic translator. If you want to bridge cultures, learn a language.



  • I want 2 of 'em and start 'em off translating back & forth in a feedback loop, as in the two famous examples:

    input: "the spirit is willing but the flesh is weak" -> russian -> back to english -> "the vodka is good but the meat is rotton"

    input: "out of sight, out of mind" -> russian -> back to english -> "blind idiot"

    etc.

  • I'm not sure how much background you have on this subject, but I lived in Taiwan for two years and have an Asian Studies Minor, so I know a little bit about Taiwan vs Japan. Most old Taiwanese that actually were alive when Japan ruled Taiwan like the Japanese. Japan did a lot to help Taiwan because they expected it be always be a part of the Japanese Empire. Now Korea (and Manchuria) on the other hand is a different matter entirely. Japan totally raped Korea of timber and any other natural resource they could because they knew they couldn't keep Korea forever.
    Okay, back on topic: I really want to check this thing out at Comdex and see how good it's Chinese translation is!!
  • An excellent, very interesting book on the subject of translation (not how-to, but about the kinds of issues that come up and some very impressive examples of translating lots of things considered to be untranslatable) is "Le Ton Beau de Marot", by Douglas R. Hofstadter, (author of Godel, Escher, Bach, an Eternal Golden Braid, another excellent book.)

    This book talks some about machine translation, but mostly about translation in general. Highly recommended to anybody interested in languages & translation.


  • Machines such as this will only serve a point when they can translate the one and only true dialect that actually has an impact on this world.... ebonics.

    Until then, I cannot envision myself purchasing one :) (Unless it does ebonics+binary)
  • I'd like to take it, reverse the mic. and the speaker, and add a borkification engine. Voila! Everyone sounds like the sweedish chef.

    Company meetings would suddenly be a source of amusement.
  • If you set up the speakers of one such device to the mic of a second, and visa versa, and set up the software so the translation path looks such:

    Device 1: English->LanguageX

    Device 2: LanguageX->English

    How many times through the feedback loop do you think it would take for the output to become totally incomprehensible in either langauge???

    Anyone going to Comdex? Try it!!!

  • Ghoti only works when spelled...if you /say/ it (goti), the translator will probably think you mean goatee.
  • The speed can't be that high. You've got to wait until the end of a sentence -- and possibly longer, to get a translation. I imagine it'll be more like listening to someone translate in one ear, much like is already done at international meetings, the U.N....

    Still. First step! Ya' have to step somewhere to get anywhere.
  • by copito ( 1846 )
    Thanks, that was a funny thread. I think there should be a special hall of fame for funny threads. My vote goes to a decent first post [slashdot.org]. Actually a lot of the posts in that article are pretty funny.
    --
  • Isn't the Tower of Babel a Biblical story, where the tower allowed everyone to understand each others' languages? I thought this was bad, leading to mass wars and such. Hmmm, hope this device isn't a tool of Satan. What OS did you say it ran? :)
  • Actually the tower of babel involved people *not* being able to understand each other cuz God gave them all different languages.

    Micah McCurdy
  • I met some of their marketing folks at a recent tech show in Maine (yeah, in Maine!). I talked with them for 20 minutes about some things and I got a sweatshirt, but I also talked to them about many other things. They have a really cool tape device - well digital recorder - which plugs in with their Voice Recognition software. It's very cool.

    Their Voice recognition is pretty decent, but I haven't played around with it enough to figure out how to add workds. I was up and runniing with the Voice Rec. software in about 20 minutes. 7 minutes install, 13 minutes Recognition testing... or whatever they call it 'signing in' I think.

    I've used the program while on irc some, and while dictating some documents. It's all windows based, but they have made some serious strides. I like it. They're promising Mac versions of things soon, but had no plans for linux, (at the time I talked to them).

    Some other things we talked about are covered in a non-disclosure, so I can't mention those... sorry.

    They're cool people... and their marketing drones are freindly, relatively technically savvy people.

    yacko

  • It was that virus that Enkidu wrote...
  • Well, the advantage of having one box that translates all the way to the target language is that you don't have to persuade other people (presumably without speaking their language) to either use the other box that you carry around with you or else to go out and buy one for themselves.
  • excellent boot? somebody needs to check their translator... ;)

    Might I suggest this translation resource? Site O' Joel [geocities.com] is a good place to translate to/from just about any language (no Somoli or Malay just yet), but it is geared toward English speakers. Oh, and its just a links page to lots of online resources, etc., but its a good starting place. Don't forget to check out the world's first ever English to Canadian [geocities.com] translator while you're there.
  • How'd you measure that head start? Not by number of fluent speakers surely. Part of the idea is to choose something that is _nobodys_ first language. Kind of levels the playing ground. Unfortunately lots of folks don't really like level playing grounds. Especially those who have the current advantage.
  • by copito ( 1846 ) on Thursday November 11, 1999 @12:27PM (#1541659)
    In case it matters, I am an American, I speak very poor Spanish dispite having a wife who is a native Spanish speaker. I spent 2 years at an international school, the Armand Hammer United World College [k12.nm.us], and have spent about 3 months of my life outside the US on various occasions. I use "Americans" in this comment to refer to United Statesians alone, and not to other inhabitants of the Americas.

    Clearly geography is part of the answer, European countries are much smaller and more integrated with their neighbors than the US. Near the US/Mexico border, there are a large number of English speaking people that speak at least enough Spanish to conduct a simple consumer transaction. As you note, the only good way of learning a language is by being immersed in the language and culture.

    Dominance is another part of the answer. In a world that is dominated by English speaking powers, particularily in economics and entertainment, most people an English speaker interacts with will have a working command of English. I suspect that when French (the original lingua franca) was dominant in diplomatic circles, that there was a similar lassitude on the part of French speakers.

    However, neither of these factors explain the aggressiveness with which Americans are monolingual. In California, where there is a large population of native Spanish speakers, bilinugal education has been banned in public schools. Elsewhere, language education for childern is half-hearted at best, if it exists at all. As one of my Spanish professors told me, the stated purpose of most elementary foreign language education in the US is to assist in the teaching of English grammar and vocabulary, not to teach for fluency.

    I think that the reason for the resistance to language education comes down to xenophobia and racism. Americans fear cultural encroachment, particularily by an increasingly large hispanic population. (This is of course ironic considering the cultural encroachment on the rest of the world by American culture, but Americans, as a rule, are poor connoisseurs of irony). Language is a particularly feared element of this cultural encroachment since language is so central to culture, and conversely a shared language connotes a certain degree of shared culture. There is of course a large measure of racism encapsulated in this fear. If Americans did not feel that immigrants were inferior, we would welcome their cultural and linguistic contributions.

    As with any prejudice, there are legitimate fears as well as ignorant ones. Since preserving language is essential for preserving culture, a multi-lingual society is a more multi-cultural one, and multi-cultural societies, notwithstanding their benefits, are more suceptible to internal conflict.

    In the midst of all this racism, xenophobia, and legitimate desire for a unifying national identity, the majority opinion is that immigrants should assimilate and learn English. This opinion is in fact shared by many if not most immigrants. Unfortunately, in the push to teach everybody English we come to the idea that not only should all immigrants learn English, but that the whole world should learn English.

    So what's the solution? I wish I knew. If you come up with a solution that doesn't involve a gun I'll vote for you.
    --
  • That's my question too.... it's nice that the thing can translate, but can it keep up with the speed of people talking, and can it also translate more than just rote. I know from my experience with foreign languages that word order and sentance structure is best done sentance by sentance (if not paragraph by paragraph) rather than word by word. Places like Babelfish do a few phrases, but the user still has to fix a lot of things. Anyone have info on this? [altavista.com]

    NIVRAM
  • I know, I know, I should shut up...

    But it was /Hungarian/, not Bulgarian.

    "I wheel not by zis tobacconist, it is scratched!"

    "Please föööndle my buttocks."
    "Ah, yes. Two blocks down, and left at the light."

    And so on...

    No! Don't click me! [surf.to]

  • by Anonymous Coward
    I am keenly interested in this development, but not for its ability to translate languages. I was born deaf, and since I deal with mostly hearing people in my professional life, I have always wished I could have a little device that would print out what people are saying to me. Having the words projected on my glasses would make me live in a subtitled world, just like a foreign movie. Most of my co-workers know enough sign language from my insistence that they learn at least a few words and I can read lips with proficiency, so life is good in my happy world. Unfortuately, it's a big company and I must deal with many people in my day to day activities on the job. Current technology that requires software to be trained isn't practical for speech to text applications. The glasses idea has been tried by others. One interesting project I saw was a tiny LED that sat in the corner of your glasses, kind of like the permanent logo on TV programs. The LED didn't print out words, but merely glyphs for each phoneme it interpreted. I am not sure how it fared in the lab because the last I heard of it was about ten years ago. I did once participate in an experiment at the local University in which a very nice young lady strapped a sleeve to my forearm, and on the sleeve was a matrix of 64 motors which would buzz against my skin. The idea was that by recognizing patterns I would be able to recognize translated speech. It worked just about as well as you would expect but it did seem to improve my golf swing. At any rate, a wearable computer capable of translating spoken words into text without any training in any given situation would be the payoff for people such as myself looking to get ahead in their careers in an imperfect, non-sign language speaking world. I am looking forward to learning about this product and I hope it can be modified with ease to applicaitons such as mine. -Derrick
  • Or be prepared for a quizzical expression.
    --
  • by the eric conspiracy ( 20178 ) on Thursday November 11, 1999 @12:39PM (#1541664)
    As Americans, we have this belief that everybody should learn English to talk to us.

    I learned a foreign language in school. I found it to be a nearly complete waste of time. The fact of the matter is that the vast majority of Americans do not live in a place where they are exposed to non English speaking people. With the exception of some parts of Los Angeles and Miami (that most people would generally not consider desirable holiday destinations) English is the first language in the US. The geography of the US is the determinant here. With time and lack of use the hard won skills of speaking atrophy and are forgotten. Sure, there is some residual benefit, but without practice the skills are lost, and surprisingly quickly.

    If I lived in Europe or some other part of the world where I was exposed to people whose native language was not English as a matter of routine, I am sure I would have found it worthwhile to develop my linguistic skills further. But the fact of the matter is that I, like almost all Americans just do not find exposure to other languages a part of daily life.

    Americans are mono-lingual for the sole reason that there are very few times in their life where another language would be useful. If the US were parceled up into a bunch of states where each state spoke a different language, you can bet that there would be a lot more interest in being polylingual. But it just isn't so. We aren't like Europe where a large country (say France) is the size of one of our states.

    The concept that all Americans believing that others should learn English to speak to us is ridiculous. First it is a stupid stereotype, and second it ignores that it is just a fact that most Americans know only English because they rarely meet people who are not native English speakers.

    You are fluent in Russian. Fine. Do you know how many people I have met in my life (I am 49) whose native language was Russian? One. An emigrant who was the fiance of a business acquaintance. How am I supposed to justify spending years of my life learning a skill that would be used maybe for 3 hours over the course of my life?

    In Europe you may learn Italian (or at least enough to perform the daily tasks) if you plan to visit Italy on vacation. But after having made the effort you have a skill that you can use frequently. Going to Italy is a two hour drive, which you may do every third summer. You may in fact meet visiting Italians frequently in your home city. How many Americans visit Italy on a frequent basis? Not many. Have you ever met an Italian tourist in the US? I haven't. Ditto German. If you go stand on line in the Louvre you are likely to meet more Germans than French. Go stand in line at the MFA in Boston and you find that there are few tourists from other countries.

    It just does not pay for most Americans to become fluent in other languages. If they make the effort they usually find that they never have a chance to use the skills and find that they have wasted their time.

    What I want to know is how many Europeans speak non-European languages. Geographically it is no different for a European to be ignorant of Persian than it is for an American to be ignorant of German. I would bet that the answer is that there are no more Europeans that speak Persian than there are Americans.

    As far as mechanical translators being useless, well I will agree that in their current incarnation they are in fact useless. But then again people never thought that it would be possible to build a machine that could beat the world champion at chess, either. Who knows what the future of mechanical translation is? Another decade or two of Moore's law and careful programming and you might find that the issues of idiom and context are solved.

  • However, neither of these factors explain the aggressiveness with which Americans are monolingual. In California, where there is a large population of native
    > Spanish speakers, bilinugal education has been > banned in public schools. Elsewhere, language
    > education for childern is half-hearted at best, > if it exists at all.

    In part that's because so called 'bilingual' classes in some areas (not all of them) were never getting around to teaching children english at all, leaving the kids at a major disadvantage after they left school.

    Not to say that all bilingual education is bad, but in this case, there was a reason behind the 'ban'.


    > As one of my Spanish professors told me, the > stated purpose of most elementary foreign > language education in the US is to assist in > the teaching of English grammar and vocabulary, > not to teach for fluency.

    This is true. More, all my school-language (latin, german and japanese) classes put a lot more emphasis on grammer than was useful, if I just wnated to *speak* the language (well, with latin maybe this is not as much of a concern) -- OTOH, I just took a japanese class that emphasized *communication* over technical correctness, and discovered that I *can* learn another language (for a long tiem I just thought I was a complete idiot with languages or something, because remembering stuff by rote does not come easily to me)
  • I'd say the head start should be measured by the number of people who use a given language as thier second (or third,fourth etc.) language. Another thing that should be taken into account is the need to communicate with others who don't share your own language - in other words peasents in the mountains of china have no real need to communicate with people who don't speak chinese, so we don't really need to take them into account. I think that under these critera English probably does have a significant lead.
    Bradley
  • Its really a pendant attached to a 100 lbs cage on wheels containing a 4-foot linguist on loan from a failing European power.


    "The worst part is the number of Stephen Hawking prank calls will rise exponentialy."


  • I suggest lojban [lojban.org]...


  • Not quite. Pronounce the 'gh' as in 'enough', 'o' as in 'women', 'ti' as in 'nation'.

    Incidentally, one of the first things I tried when I got S.A.M. (Software-Automated Mouth) for ye old Commodore 64 was 'ghoti', which it pronounced 'gosh'.

    Also incidentally, I think one of the words listed is misspelled. 'floccinaucinihilipilification', the act of judging a bit of information to be utterly worthless, according to the (dis)honorable Cecil Adams in _More of the Straight Dope_.

  • for some reason I seem to recall a piece of software translating/pronouncing/interpreting the word ghoti "correctly" (ie. fish) as a result of some smart ass programmer sticking it in as a special case.

    anyone else remember something like this? fantastic work by the coder anyway.

  • nice

    last year I thought it'd be fun to test some text-to-speech systems by sticking the output through a speech recogniser, and comparing to the original input.

    needless to say there's still a lot of work to do on TTS

  • Star Trek here we come......but when will it be the size of a communicator. I am there......
  • does it fit in your ear?
  • Have that gotten rid of that -horrible- computer synth voice we've had since the days of War Games??? I can see it now, a Russian diplomat walks up to an American diplomat and asks the question that ends the world.
    "Would you like to play a game?"
  • Don't pay any attention to "accuracy" (mis)claims from manufacturers about these "translators". Believe it, these things simply cannot be any good.

    Natural language processing is not advanced enough to produce any kind of real-world acceptable translation. Especially when we are talking about realtime voice recognition + translation stuff.

    BTW, I'm a linguistics grad student. Not that that makes me be right in anything I say above, but at least I'm not speaking out of gross ignorance.

    ---

  • Anything that translates to and from so many languages must have some meta-language that it works with internally.

    Well, this is either trivially true or patently false, depending on what you meant with "language". If you mean something similar to a natural language (like Esperanto as you suggest below) then it is false. You know, parsing and interpreting a sentence in a natural language (a task that no one knows how to do as of now) is more than complicated enough of a task; if you have managed todo this, you obviously want to store the information in some "language" that's easy to decode.

    Schematically, what a translator should do is parse natural language utterances, derive from them semantic representations, and then generate from those as semantically close as feasible utterances in the target language.

    What a "semantic representation" should precisely look like no one knows; there is really A LOT involved in it. Pick up an introductory textbook in Semantics and Pragmatics (like Chierchia and McConnell-Ginet, Meaning and Grammar, 1990) and look through it to get an idea.

    The short story is that a semantic representation for a sentence has to include tons of stuff that might not be present in the sentence itself. For a very simple example, the English sentence "I'm handsome" is lacking some information that its semantic representation should have; namely, whether the sentence should be intepreted as meaning "I'm handsome now" or as meaning "I'm always handsome". This can only be determined from context (what the hearer knows or can infer about the speaker). Yet if you cannot make a correct assumption as to which of these meanings to assign to particular utterance of this sentence, you cannot correctly translate it into Spanish. Because you have to choose between "Yo estoy guapo" and "Yo soy guapo". These sentences each correspond to one of the two possible meanings of "I'm old" mentioned above.

    So what is that Meta-language, and can we figure out how to turn it into something that could be used as the universal language? Perhaps all communications should be translated into that on public channels.

    Pure fiction. We simply don't know what a semantic representation looks like, but we do know that it is bound to be enormously complicated, too much for it to be useful (or for us to use it consciously with any significant skill, for that matter).

    ---

  • >You could go as far back as the Bible, if you wanted. In the book of Acts, on the Day of Pentecost, St. Peter et al were speaking in their own language and their words were perfectly understandable to "all in their own tongue". But that one's usually chalked up to "Cool Miracle", not "Cool Technology" status.

    [stephenson]*mumbles something about it not being cool at all when virii hijack a religion*[/stephenson]

    What, POT mode STILL eats tags? Grr...


    --
    "HORSE."
  • What I want to know is how many Europeans speak non-European languages. I would bet that the answer is that there are no more Europeans that speak Persian than there are Americans.

    You'd be surprised. Europe is a pretty popular destination for political refugees from that region, so there's more exposure..
    -John
  • This is going to be an incredible help to so many people. I wonder how many people will lose their jobs.
  • You need to know both languages to judge machine translation. So far I've not been impressed, most of the time machine translation is more like instant humor generation. The only thing I tried that was close to ok was a simple travel description, with a few very odd words sprinkled in (this was babelfish, german to english). I never would use it to translate into a language I don't know. Some instant humor for the Germans: Sie müssen beide Sprachen kennen, um maschinelle Übersetzung zu beurteilen. Bis jetzt bin ich nicht, die meisten der Zeitmaschinellen Übersetzung bin mehr wie sofortiges Stimmungerzeugung beeindruckt worden. Die einzige Sache, die ich versuchte, der nah an O.K. war eine einfache Spielraumbeschreibung war, wenn einige sehr ungerade Wörter innen besprüht sind (dieses war das babelfish, Deutsch-Englisches). Ich nie würde sie verwenden, um in eine Sprache zu übersetzen, die ich nicht kenne.
  • Just like pocket translators, every first year foreign language student is going to want one so he can learn how to say all the naughty words. At least now the teacher won't have to tell them that they are pronouncing their swear words wrong.

    Even if travellers use them, I doubt that they will be taken seriously. Can you imagine someone walking up to you with a little box, fumbling to make sure it hears him while he talks into it, then thrusting it at you for it to speak something that is more than likely SOMEWHAT recognizable as English (-or insert another language of preference here)? I would just laugh.
    --
  • I don't know if a computer will _ever_ be able to replace the ability to actually comprehend a language. I mean, I use bablefish at least once a day, but I would still have no idea what the languages are saying if I knew nothing about the language's idioms and sayings.

    What I don't understand is: why is it so difficult to learn another language? I mean, isn't there a stat like only 10% of a language is used 90% of the time? And in all indo-european languages, like 9 out of that 10 percent is all the same. So why can't I learn French in 1 day? I'm trying to cram as many languages into my brain as I can (working on French, Hungarian, and Latvian, wish me luck), and doing it without any formal instruction.

    I came to France 6 months ago savoir _aucune_ de francais, and now I can almost speak it. I don't know if that's slow or fast, but why did it take me 6 months to learn the few hundred main words of this language?

    If there are any linguist /. readers, please email me. This subject facinates me and I have no education on it.

    -davek

  • it was my impression that this was a voice-to-voice system, so you must speak into it. If you spoke ghoti (and didn't say it "fish") then it would probably think "goatee". Of course if it is text-to-voice, then ghoti would be a screwball ;)
  • so i see if your traveling, but if your hosting an oversea's confrence call consider this; while it has been said that it probaly would be lacking in accuracy, if you made a box that was only (laungauge)-to-esperanto, you could make things a lot more accurate. so, your company is in Norway using a norse-to-esperanto; and the other company is in japan with a japanese-to-esperanto. The result would be a more accurate translation? +in a company-to-company situation, the hassle/cost is shared and not so much of a burden. you can buy your (laungauge)-to-esperanto translator and pay not for all the rest included (long term: more laungauges total).
  • Imagine feeding the output from a 95% accurate speech recognition system though babelfish.

    Check out the past story [slashdot.org] about the neural network that recognizes speech better than humans.

    And on a similar subject, I think it'd be interesting to see neural networks applied to translation - they could possibly be trained to recognize idioms, produce more meaningful output, etc.

  • Now, how are we going to email you, when you don't give your email address on /. and when your given web site also shows no trace of a "davek"???

    Doh!

    Anyway, a large part of the answer is that you need to do more than just "learn" a few hundred words. You could probably memorize them in a week of concentrated study (or less). But that's not enough. You need to develop conditioned reflexes of associating the meaning with the sound (and vice versa) nearly instantly and unconsciously.

    Developing conditioned reflexes takes a long time. Compare this with learning a martial art, or gymnastics, or learning dozens of complicated dance routines -- all of these examples could easily take 6 months or even years.

    So we shouldn't be surprised at the length of time that it takes to develop really good conditioned reflexes in language, either.

    (Pity no one will ever see this comment, since I'm posting 6 days after the discussion!)

A sine curve goes off to infinity, or at least the end of the blackboard. -- Prof. Steiner

Working...