Forgot your password?
typodupeerror
Google AI Technology

Google Improves Android Translator To Battle Siri 185

Posted by samzenpus
from the let-your-phone-do-the-talking dept.
judgecorp writes "Google Translate for Android, the mobile version of Google's machine translation software, now translates speech back and forth between 14 languages, the company claims. Earlier this year the company added Conversation Mode, which lets users to translate chats between English and Spanish. Now Google has made the tool available from Android 2.2 handsets and later in Brazilian Portuguese, Czech, Dutch, French, German, Italian, Japanese, Korean, Mandarin Chinese, Polish, Russian and Turkish. The arrival of Siri on the iPhone could spark serious competition in translation systems on phones."
This discussion has been archived. No new comments can be posted.

Google Improves Android Translator To Battle Siri

Comments Filter:
  • by Oswald McWeany (2428506) on Monday October 17, 2011 @09:34AM (#37738178)

    How long until this technology spreads?

    If I open the fridge will it remind me that the premixed salad I intended to eat has been sitting there for two weeks and it will go bad today if I don't eat it.

    Can I then ask my fridge what shelf I put the salad on because I don't remember- and it will tell me.

    Will my front door greet me and ask how my day was- and be compassionate to me when I say it sucked. Could I ask my front door if it has seen anything suspicious- and change the welcome message for salesmen to "go away"?

    Will my dog bark to me when I come home- and come running when I say "walkies"?

    • by 0100010001010011 (652467) on Monday October 17, 2011 @09:46AM (#37738336)

      Dangerous times lay ahead when asking your appliances anything about tossed salad.

    • by shadowfaxcrx (1736978) on Monday October 17, 2011 @09:47AM (#37738346)

      Yes.

    • by TheDarkMaster (1292526) on Monday October 17, 2011 @10:54AM (#37739170)
      ... until this tech can correctly translate from Brazilian to English?

      As an example, this sentence in Portuguese:

      "Vamos evitar o uso de papel, gastar papel implica em gastar árvores"

      Google translates as:

      "We avoid the use of paper, wasting paper implies spending trees"

      Here we have some problems of grammar, changed words for no reason and wrong use of future. A more correct translation is*:

      "We will avoid the use of paper, spending paper implies spending trees" br>
      * Note: Is not a "exact" translation. English is too simple to pass the same idea in the same way as using Portuguese.
      • by jfengel (409917) on Monday October 17, 2011 @11:29AM (#37739648) Homepage Journal

        Thing is, I got the gist of it. It may not be a great translation, but if I'm in Brazil and the options are (a) that translation, and (b) no translation, I'm a lot better off with (a) than (b).

        You could call it 80% of the way there, in four decades of work. The remaining 20% will probably take another four decades, at least. But at 80% we've reached something that is frequently useful.

        • by JasterBobaMereel (1102861) on Monday October 17, 2011 @12:25PM (#37740538)

          Getting the basic translation without context is easy, getting the basic translation with some syntactical context is hard, getting the full translation with complete context means you understand both languages

          Currently machine translation does not understand either language, it just has a set of rules to translate with context, the more context the better it works but it will still only get you 80-90% of the way there ...the last 10% is the difference between good enough translation (which we have now) and near perfect translation which professional human translators do...and can mean the difference between misunderstanding and comprehension ...

          Siri was designed to be a command interface - it copes quite well in a fairly small domain, Google translate does a fairly good job across the board

      • by Rei (128717) on Monday October 17, 2011 @12:31PM (#37740630) Homepage

        Hey, at least the Portuguese translation isn't full of *deliberate* jokes like the Icelandic translation. They've fixed a number of them (like translating "Sigur Rós" as "Foo Fighters", translating "Where is the bathroom?" as "Talarðu ensku?" (Do you speak English?), translating "My hovercraft is full of eels" as "Láttu mig í friði!" (Leave me alone!), etc), but there's still a *ton* in there. Someone had a lot of fun exploiting how it builds its translation database. :P I guess the fewer the speakers, the easier it is to rig. And even where it's not rigged with fake translations, it often likes to pretend that whole words or phrases -- very common ones, at that! -- aren't in the sentence and just leaves them out of the translation.

        There are uses for Google Translate. Getting a *good* translation isn't one, but getting a rough idea quickly is. Given sufficient time (and a good dictionary for the words/phrases I don't know), I can translate pretty much anything from Icelandic -- but I don't always have the aforementioned "sufficient time". There's always time for Google.

      • by Pope (17780) on Monday October 17, 2011 @01:01PM (#37741058)

        Except "spending paper" isn't an English phrase, so "wasting paper" is the better idiom.

        • by TheDarkMaster (1292526) on Monday October 17, 2011 @02:20PM (#37742118)
          Maybe, but in Portuguese you loses the meaning... "Wasting paper" translates to "jogar fora papel", "jogar fora papel" is not the same thing as "gastar papel".
          • by Anthony Mouse (1927662) on Monday October 17, 2011 @07:40PM (#37745300)

            Then you're complaining about it in the wrong direction. The translation from Portuguese to English is correct. Translating "wasting paper" to "jogar fora papel" is what is wrong, because it is a literal translation of an idiom.

            What it (and you) get wrong is to translate "gastar árvores" as "spending trees" when it should probably be "killing trees" -- but unless "gastar árvores" is an idiom in Portuguese that seems understandable, because it is the literal translation. It just doesn't come across properly in English because "spending" is not generally something you do to a tree (or paper).

            • by Carnildo (712617) on Monday October 17, 2011 @08:30PM (#37745678) Homepage Journal

              What it (and you) get wrong is to translate "gastar Ãrvores" as "spending trees" when it should probably be "killing trees" -- but unless "gastar Ãrvores" is an idiom in Portuguese that seems understandable, because it is the literal translation. It just doesn't come across properly in English because "spending" is not generally something you do to a tree (or paper).

              When you're translating the sentence, you want to preserve the parallel construction in the second half of the sentence: whatever you're doing to the paper is what you're doing to the trees. It's a stylistic technique that increases the strength of the sentence, and is an example of why good translation requires the translator to be at least as good a writer as the original author.

            • by TheDarkMaster (1292526) on Monday October 17, 2011 @10:58PM (#37746490)
              Well, on brazilian portuguese "wasting" (jogar fora) and "spending" (gastar) is not the same thing. "Jogar fora" (wasting) on brazilian is to throw away something, throw on junk, put on garbage, etc. While the word "gastar" (spending) means using your money to buy something or using resources (the tree) to do something (making paper). Extending the original sentence she would like as "We will avoid use of paper, spending paper on printers need us to spend trees to make paper".

              You uses "wasting" (jogar fora) if the idea is "you are wasting resources, putting then on the garbage" on portuguese, but if the idea is more or less "you are using resources on a thing that is not waste", you uses the "spending" (gastando).
    • by tehcyder (746570) on Monday October 17, 2011 @11:54AM (#37740062) Journal

      If I open the fridge will it remind me that the premixed salad I intended to eat has been sitting there for two weeks and it will go bad today if I don't eat it.

      Can I then ask my fridge what shelf I put the salad on because I don't remember- and it will tell me.

      You're either got the world's biggest fridge or you're the laziest, stupidest person on the planet.

    • by dna_(c)(tm)(r) (618003) on Monday October 17, 2011 @12:30PM (#37740616)

      "My hovercraft is full of eels." [citation [wikipedia.org]]

    • by nospam007 (722110) * on Monday October 17, 2011 @01:14PM (#37741234)

      "Can I then ask my fridge what shelf I put the salad on because I don't remember- and it will tell me."

      Nice. When we finally had some quiet time during commute, because the teens are texting instead of phoning, now they'll talk with their _phone_ instead of just clicking the damned weather app.

  • by Lord Satri (609291) <(moc.liamg) (ta) (xuorelerdnaxela)> on Monday October 17, 2011 @09:35AM (#37738182) Homepage Journal

    Siri doesn't do translations, it's more of an advanced voice recognition tool. Am I wrong? This would mean that at the moment, Apple's Siri and Google Translation would have two different strengths; Siri: usable natural language voice recognition (at least that's how they sell it) and Google Tranlation, well, multi-language translations.

    • by monkeyhybrid (1677192) on Monday October 17, 2011 @09:50AM (#37738372)
      You are correct. I'm sure Siri will gain more functionality in the future but as it stands now, Siri should be compared to Voice Search [android.com] and other third party voice command apps.

      Google Translate does translate.

      Siri (on iPhone) and Voice Search (on Android) handle voice commands and interaction.
    • by goombah99 (560566) on Monday October 17, 2011 @09:52AM (#37738402)

      If you can think back to your last conversation with a waiter in a country whose language you did not speak then you know that Translation requires fairly high accuracy to be perceived as being good. Commands don't. Siri is trying to get the jist of what you mean and act on it.

    • by adamstew (909658) on Monday October 17, 2011 @09:52AM (#37738406)

      Siri isn't even just voice recognition. It uses voice recognition as the input for it to perform tasks or lookup trivia type questions for you. Things like "schedule a meeting with my boss tomorrow at 10 about next year's budget". That's all you need to say, and Siri will create a meeting invite in Exchange, inviting my boss for a meeting at 10... it'll title the meeting "Next Year's Budget".

      You can use it for voice dictation... Just about every text input field, now has a microphone that simply lets you say what you were going to type... and the voice recognition is VERY good.

      So while google is using voice recognition, it is using it for two different applications. It's like saying Microsoft Word is competing with Adobe Photoshop because they both use the mouse as an input device. You can't even use Siri for translation. But it wouldn't surprise me if Apple added a similar type of translation feature to Siri... although I doubt it would show up in a conversationalist way... More like "translate 'where is the nearest hospital' in to french".

      • by Terrasque (796014) on Monday October 17, 2011 @10:21AM (#37738736) Homepage Journal

        You can use it for voice dictation... Just about every text input field, now has a microphone that simply lets you say what you were going to type... and the voice recognition is VERY good.

        And that have been on Android since v1.6, actually.

        • by lpp (115405) on Monday October 17, 2011 @10:48AM (#37739078) Homepage Journal

          My understanding (disclaimer: I do not have a 4S, I'm solely relying on reviews) is that Siri handles natural speech much better than the speech recognition found on Android devices. By natural, I mean regular conversational tone and flow, as opposed to the more robotic method where you must. pause. on. each. word. in order to give the translator the chance to note word breaks and parse the sentence structure. So it's not that Siri does it and Android doesn't. It's that Siri seems to do it better, working for the user instead of the user working for the computer.

        • by dogmatixpsych (786818) on Monday October 17, 2011 @01:54PM (#37741818) Homepage Journal
          My limited experience with Google's voice recognition on my Android phone is that it is okay but nothing special. Actually, it's far more annoying than just typing what I want. I'll be saying something and it will think I'm done talking and stop listening to me. That's mainly why I haven't used it more - it's so buggy that it's not worth the effort. Maybe it's just my phone or my version of Android (2.2) or my technical ineptitude but voice recognition in Android needs a lot of work to get to where Siri is now.
    • by brunokummel (664267) on Monday October 17, 2011 @09:54AM (#37738422) Journal

      ...... Siri: usable natural language voice recognition (at least that's how they sell it) .....

      Japanese people beg to differ... :D [youtube.com]

    • by wombatmobile (623057) on Monday October 17, 2011 @09:54AM (#37738430)

      Siri doesn't do translations, it's more of an advanced voice recognition tool.

      Siri is a user shell. It provides an interface to the operating system. Voice recognition is one part of the technology; a vital part. The other part is the logic Siri uses to process the words it hears as commands. It translates spoken natural language statements into operating system calls.

    • by MobileTatsu-NJG (946591) on Monday October 17, 2011 @09:55AM (#37738438)

      You aren't wrong, Siri is not a translator. However, Siri's grasp of context is, hypothetically speaking at least, and important step for translation. I'll give you an example: Before I did a sync of my phone, Siri didn't understand my cat's name. Afterwards, now when I mention my cat, it spells the name correctly and even capitalizes it as a proper noun. I was seriously shocked that it picked up on that. For that reason, I think it can tell the difference between may and May.

      Anyway, I'm getting a little off-track here. Yes, you're right, it has nothing to do with translation. However, I can totally see why people are starting to dream of that possibility coming awfully quick.

      • by Anthony Mouse (1927662) on Monday October 17, 2011 @08:26PM (#37745658)

        However, Siri's grasp of context is, hypothetically speaking at least, and important step for translation.

        I think this is the key point here, especially when you consider it from the other way around: If you have a translator, you have something that inherently needs to understand the meaning of language so that it can translate it properly.

        So let's say you do something in the nature of defining a new language as a series of functions. "Call Bob" means "telephone_connect(Bob)" etc. Now you can translate from English (or whatever other language the user speaks -- or types) into some kind of 'action language' that the device can execute directly. The ability of the translator to detect context and change the meaning (and therefore the action) is a big feature -- the better a translator it is, the better it understands what you mean when you tell it to do something. You end up with a natural-to-machine-language translator, but in a way that could actually be usefully implemented -- because you take the code already written (like making a phone call) and make it available to the translator as part of the language being translated to.

        You still end up with the unavoidable problems (e.g. confusing sound-alike words that both fit the context, or linguistic ambiguities like 'bob took a picture of alice on the patio'), but it makes a big step toward an interesting result.

    • by poofmeisterp (650750) on Monday October 17, 2011 @10:16AM (#37738684) Journal

      Heh.

      What do they call the process where you say that you're better than the competition in a VERY loosely-related type of product?

      e.g. "In the news today, Proctor and Gamble, Inc. has announced that it plans to compete with the largest trucking companies. P&G claims that its delivery processes are faster than all other trucking companies."
      "Well, they USE trucking companies, so their food must be the best if they are competing with the other trucking companies, right?"

      • by macs4all (973270) on Monday October 17, 2011 @12:29PM (#37740596)

        Heh.

        What do they call the process where you say that you're better than the competition in a VERY loosely-related type of product?

        e.g. "In the news today, Proctor and Gamble, Inc. has announced that it plans to compete with the largest trucking companies. P&G claims that its delivery processes are faster than all other trucking companies." "Well, they USE trucking companies, so their food must be the best if they are competing with the other trucking companies, right?"

        I think they call that Strawman Advertising...

        You are exactly right. Perfect analogy, too.

    • by vawwyakr (1992390) on Monday October 17, 2011 @10:38AM (#37738968)
      My assumption is that Translate is used by other Apps to do voice recognition. I have Vlingo on my Android phone which is very Siri like (it does dictation, search and other such things, I had a phone off with my IPhone 4S using friend on Friday and other than being a bit faster they seemed pretty similar).
      • by macs4all (973270) on Monday October 17, 2011 @12:32PM (#37740648)

        My assumption is that Translate is used by other Apps to do voice recognition. I have Vlingo on my Android phone which is very Siri like (it does dictation, search and other such things, I had a phone off with my IPhone 4S using friend on Friday and other than being a bit faster they seemed pretty similar).

        So, you got together with a friend and phoned-off?

    • by z_gringo (452163) <z_gringo@hot m a il.com> on Monday October 17, 2011 @11:03AM (#37739296)
      Not only does Siri not do translations, the command / control functions that she is supposed to do, are extremely limited. Siri, at least for me, fails to work most of the time even for the incredibly simple tasks that I have tried, such as calling contacts and playing certain songs.

      For example, Siri can set the alarm for 6:00am, but if I want that alarm to be set for every weekday, that task is too complex for her.

      Siri can play songs, provided that your song list is filled with popular songs in english by artists with English sounding names, but if you are using Siri in English and happen to have songs in German by German artist, then Siri is useless. Ditto for making calls to contacts whose names aren't common names for the language you are using Siri in.

      Siri actually doesn't seem to be much better than IBMs voice recognition software that I used nearly 20 years ago.

      I can't see why anyone is remotely impressed with Siri, it is just a useless waste of time as far as I can tell. Google translate, on the other hand seems really impressive with being able to recognize the speech and translate it to another language.
    • by Quiet_Desperation (858215) on Monday October 17, 2011 @11:24AM (#37739592)

      You are correct. The whole article and comparison makes zero sense.

  • by Pieroxy (222434) on Monday October 17, 2011 @09:37AM (#37738202) Homepage

    Why the catchy headline "To Battle Siri"? Why wouldn't it be just for "Improving Android Translator"?

  • by Grizzley9 (1407005) on Monday October 17, 2011 @09:42AM (#37738280)
    Does Siri even do translation? Every time I've asked it to translate an English word into Spanish it says it doesn't know what I mean. This is not "battling" Siri at all. Catchy key-word title to get more clicks is what I see.
  • by SebZero (1051264) on Monday October 17, 2011 @09:44AM (#37738306)

    The conversation mode is in alpha and it's intermittently very good or very bad (a complete hoot)!

    I'm bilingual and visited my mother with a, "mum come have a go at this" - 15 minutes later we gave up with tears of laughter at some of the translations.

    Then of course, we tried pieces of the "voice recognition lift" skit which has again come into relevance with the release of siri

    http://www.youtube.com/watch?v=5FFRoYhTJQQ [youtube.com]

  • by kervin (64171) on Monday October 17, 2011 @10:01AM (#37738516) Homepage

    First, the article makes no sense since Siri doesn't do translation. I guess translation doesn't "exist" yet since Apple doesn't have a product.

    Google, Nuance and Microsoft have been pushing Speech Recognition for a few years now. These companies put millions into NLP R&D ever year and are on the forefront of technology. Apple had been ignoring this space and so these companies have had great Speech Recognition and other NLP products for a while and Apple doesn't.

    Google and Microsoft are about to release the next wave of speech products ( e.g. in Android 4 and WP 8 ). These companies have NLP technology Apple hasn't even begin to tackle. Like NLP in all major world languages and across many markets ( eg. Checkout EngKoo [microsoft.com] for example )

    IOS was falling behind and Apple scrambled to purchase a Speech recognition mobile app, quickly licensed Nuance and Wolfram Alpha knowledgebase technology, and added those APIs in the operating system. They had to remove Siri from their market place.

    Marketing mentions DARPA, but just about all Speech R&D is funded in someway by DARPA. DARPA's been carrying that torch for a while now. Even the popular open source Pocket Sphinx [sourceforge.net] was made possible by partial DARPA funding.

    In short this Siri marketing push is the largest scale astroturf marketing campaign I've ever seen.

    • by Henriok (6762) on Monday October 17, 2011 @10:21AM (#37738734)
      The problem with everything you mentions is that it can be awesomely great, but fundamentally useless if no one knows that it excises at all. I didn't know Google have any speech services, nor Microsoft. I do however know that Apple has, so what Apple have done is useful for me. What Google or Microsoft has done was impossible for me to use, since it was unknown to me.
      • by Nemyst (1383049) on Monday October 17, 2011 @12:07PM (#37740262) Homepage

        If you read the post properly, you'll notice that it'll be in the next generation of WP and Android devices. It's rather logical that neither Google nor Microsoft has said anything about it thus far since they haven't announced any device running on said OS. We're expecting them momentarily, at least for ICS, but it's not there yet.

        The marketing push will start when devices are announced, assuming they have something ready.

    • IOS was falling behind and Apple scrambled to purchase a Speech recognition mobile app, quickly licensed Nuance and Wolfram Alpha knowledgebase technology, and added those APIs in the operating system. They had to remove Siri from their market place.

      Actually, Siri was bought by Apple. They do license a bunch of stuff from Nuance and probably incorporated it into Siri, but from the Siri devs, the 3GS was simply incapable of doing the necessary processing without a lot of shortcuts.

      Apple saw the Siri app, was impressed, and bought the company so they could integrate it into the OS (and have more deeper integration) as well as remove the cheats that were done in order to run smoothly on a 3GS.

      At best they added WolframAlpha to Siri's database to consult, but I'm pretty sure that was a backend integration task with WolframAlpha than Apple buying that stuff.

    • by hawaiian717 (559933) on Monday October 17, 2011 @07:20PM (#37745130) Homepage

      Apple had been ignoring this space

      Well, except for the PlainTalk [wikipedia.org] technology they introduced in 1993.

      Google and Microsoft are about to release the next wave of speech products ( e.g. in Android 4 and WP 8 ). These companies have NLP technology Apple hasn't even begin to tackle. Like NLP in all major world languages and across many markets.

      So, the technology you can get from Apple today is inferior to technology you'll be able to get from Google and Microsoft at some point in the future. This is neither surprising nor a valid argument.

      Nor do you know what Apple is working on for future releases, but it's can be said with reasonably high assurance that Siri 2.0 will be better than the Siri we have today.

    • by King_TJ (85913) on Monday October 17, 2011 @11:26PM (#37746598) Journal

      I suppose you're technically correct (although I believe Apple was dabbling in voice recognition a bit with newer iPods, before Siri was even on anyone's radar, as a way to control basic music player functions, hands-free). But IMHO, this whole "Siri popularity" thing has as much to do with the implementation as the technology itself.

      As is typical for Apple, they've taken an existing technology ... perhaps even one that's not clearly "best of breed", but found a way to integrate it so the average user will actually USE it regularly and enjoy it.

      The humorous responses Siri gives to many queries and the friendly way it handles others goes a LONG way towards hooking new users on it. I definitely don't recall any voice assistant software on Android that would give playful or creative replies to basic questions about the weather. With Siri, I might ask "How will the weather be in Hawaii on Thursday?" and get back a reply like "Bring your sunglasses! It looks like it will be warm and sunny!" At best, I'd expect a dry, literal response from the likes of Microsoft or Google.

      Also, IMO, it's no small detail to note that Apple wound up licensing Nuance and Wolfram technologies. You could do FAR worse than combining those two powerhouses for your voice recognition needs! If I had to bet, I'd put my money on the two of them doing a superior job in the speech recognition arena to Microsoft (who has their hands in as many cookies jars as possible at any given time, vs. specializing in such things).

  • by norminator (784674) on Monday October 17, 2011 @10:10AM (#37738612)

    which lets users to translate chats

    Who translated this article?

  • useless (Score:4, Insightful)

    by Blymie (231220) on Monday October 17, 2011 @10:30AM (#37738856)

    Google's voice search, and translate, and all other speech -> text products are absolutely useless. I find a very, very, very low success rate.

    Good luck if I'm driving in a car too, and the background noise adds to the difficulty.

    I know many other people that are in the same boat, but these are all locals. I wonder if other people think it's just great to have to repeat themselves 10 times, or if others do not have the same issues. I do not believe that local dialects and pronunciation is the issue, the english I hear here, seems to be the same english I might hear on Northern US news reports, on TV stations.

    However, as it sits? Useless!

    • by jfengel (409917) on Monday October 17, 2011 @11:32AM (#37739690) Homepage Journal

      I haven't used Google voice recognition in this context, but I have seen it in use on YouTube videos. I have yet to see a single case, even for simple spoken speech, where it was even describable as usable.

      Their translator generally gets the gist of things, from properly-written text. But the speech-to-text part of it is not ready for prime time, at least not for naturally spoken speech. Maybe it's better when you're speaking directly to it.

    • Re:useless (Score:4, Informative)

      by ergo98 (9391) on Monday October 17, 2011 @11:36AM (#37739762) Homepage Journal

      " I do not believe that local dialects and pronunciation is the issue"

      I have been using the voice input functionality since it came out, and have been shocked at the startling accuracy of it. It is almost never wrong, and is eminently useful for navigation, making calls (by number or by name), or for voice dictation in a message. I use it frequently and it is shockingly rare that it isn't dead on.

      I'm talking about just general voice to text, not about translate which adds another language to language issue, however Google has the voice recognition thing DOWN. I imagine there are some accents and manners of speech that present it difficulty however.

  • by mswhippingboy (754599) on Monday October 17, 2011 @11:15AM (#37739438)
    I'm looking forward to the day when both iOS and Android devices both have continuous speech recognition tied to chatter-bot apps so that we can just sit back and watch them argue amongst themselves about which is better, thereby saving ./ 80% of the conversion space.

    Then we can focus on something really important like who was more influential - Dennis Ritchie or Steve Jobs....

  • by mswhippingboy (754599) on Monday October 17, 2011 @11:36AM (#37739766)
    As an aside and part of the Stanford introduction to AI course, Peter Norvig discusses some technical aspects of how Google translate works under the covers. Pretty interesting for the nerds among us...

    http://www.youtube.com/watch?feature=player_embedded&v=sPSN0aI0PgE [youtube.com]

  • by utkonos (2104836) on Monday October 17, 2011 @10:39PM (#37746384)
    Google's speech to text capabilities are essentially trash [washingtonpost.com].

"Don't worry about people stealing your ideas. If your ideas are any good, you'll have to ram them down people's throats." -- Howard Aiken

Working...