Swype and Voice Recognition for mobile device inputting

« previous post | next post »

In late 2012, while visiting my son Tom in Dallas, I noticed that he was doing something very odd with his cell phone.  Most people enter text into their cell phone by pressing their thumbs (or their fingertip) on the letters of a small keyboard, whether virtual or actual.  But Tom was doing something altogether different:  he was sliding his finger over the glass surface of his phone and somehow, by so doing, he was able to enter text.  I was dumbfounded!  What amazed me most of all was how casual he was about it.  He'd be talking to me about something, then glance down at his cell phone, move his fingertip around on the glass, and — presto digito! — he'd have typed a message to someone and sent it off.

Now, Tom is in telecommunications and in general is a very handy guy who can fix almost anything, so I thought that what he was doing with his magic finger was not something that ordinary mortals could achieve.  I simply filed that mind-boggling experience away in the back of my head somewhere (though I subsequently did draft a preliminary Language Log post about it that got lost somewhere).

But two things happened recently that made me resolve to actually record my thoughts on the phenomenon of Tom's relaxed symbiotic relationship with his phone.  The first incident was in Prague this past summer when I observed a colleague from Taiwan entering Chinese text on her cell phone.  Normally, that would not be a big deal, because nearly all Chinese speakers I know use Hanyu Pinyin (Romanization) to enter text (as I have described in many Language Log posts), and it looks pretty much like someone entering English.  But my colleague from Taiwan, a professor in the Chinese literature department of National Taiwan University who is obviously highly literate and intensely devoted to the characters, presented a totally different appearance as she entered Chinese text into her phone with a shape-based system.

Unlike Tom's nonchalant swiping across the glass surface of his cell phone, the professor from Taiwan seemed to be attacking her poor cell phone.  OMG!  With a frenzied swishing, she raced her fingertip around in circles, bends, crosses, hooks, and all manner of other shapes.  Her whole body even got into the action.  I was breathless from watching her and almost had a heart attack for fear that she was going to collapse from the strenuous effort.  She was clearly practiced at what she was doing, yet it looked as though she were engaged in a most difficult task.

Juxtaposing Tom's smooth motions with the Taiwan professor's frenetic movements in my mind's eye, they seem to serve as a parable of the two writing systems.

The second thing that prompted me to write this post is that this year, when I went back to Dallas to celebrate the advent of the New Year with Tom and his family, I saw him talking into his cell phone, not for the purpose of making a call, but to enter text.  Once more, as I always am when I see people doing things that are totally beyond my ken, I was flabbergasted.  Somehow, the software in Tom's phone converted what he was saying into typed form.

Naturally, the technical aspects of Tom's two easy-appearing methods (called Swype and Voice Recognition) for inputting text are way beyond me, but for those who are interested, I have copied — with minimal editing — a few paragraphs that Tom kindly drew up for me to explain how they work.  For the most part, I have let Tom speak in his own voice, because it is that of a professional in the realm of telecommunications.  Allowing him to have his own voice affords us a glimpse into the inner workings, and even a sense of the future, of an industry that we all rely on but seldom comprehend.

BEGINNING OF TOM'S DESCRIPTION AND EXPLANATION

Before we do anything else, we have to understand the basics of cloud computing = any time the user is drawing computer processing power from a non-local machine.  In other words, cloud computing is invoked whenever you are using an online application and you are manipulating a remote computer.

Some people say even something as simple as checking your bank balance online is "cloud" computing, as you are obtaining information from a database on a remote computer.  However, I [Tom] personally draw some distinction between using a basic online service such as that and "cloud" computing.  To me, cloud computing means accessing very powerful computer processing ability remotely.

This is important background information for apprehending what I am about to say concerning Swype and Voice Recognition.

First of all, the creators of the programs should be revealed.  Swype is the trademarked name of a program created by a company called Nuance.  This is a great name for a company in the handheld device industry space, because there is a concerted market force to promote higher and higher interactibility with handhelds under the name of "gestures."  So we can interact with our handhelds in more complex ways within this concept of "gestures".  And we can look like mad genius Beethoven conductors if we really get our digits a-fluttering.

Swype came pre-loaded on my HTC Wildfire S mobile phone, and is probably widely available as a pre-loaded feature of any Android phone.  I do not believe Apple iPhone has any comparable function.

[VHM:  HTC is a Taiwanese company.]

Swype, as you already know, allows the user to move their digit over a series of letters on the touch screen of their mobile device and then predicts what word that string of letters should be.  The entire movement should be called a "Swype," and the word can be used as a noun or a verb.  "I'm swyping and my swype was 'food'."  Once a swype is completed, the program will present a choice of words to the user, with the highest ranked word first and highlighted.  The Swype program does have some additional intelligence built into the input method, which I know probably only the basics of, such as the ability to communicate double letters to the program by holding their finger over the letter longer (for example start with "f" slide finger over "o" and hold momentarily, then slide finger over "d" to produce word "food.")  The entire programming of the Swype application is stored locally on the mobile phone, so, it should not be called a "cloud" application.  To sum up, Swype is a great way to speed up entry of text messages, and it does endear the user to their little mobile device in a special way by eliminating the pecking feeling of entering individual letters on such tiny screens.

Link to Nuance corporation website.

Wikipedia article on Swype.

++++++++++++++++++++++++++++++++++++++++++++++++++

Next I'm going to write about Google Voice Recognition (aka Speech Recognition).

I believe the Google Speech recognition is a much more interesting phenomenon than Swype.  This is because of the non-local computing aspect (it is most definitely a "cloud" computing program).  I did a bit of internet sleuthing and found some articles about it; I will provide the relevant links to them at the end of my description.

Prelude: Mobile phones OS and makers discussion

When discussing mobile phones, it is critical to understand that Google created the Android operating system.    This can be thought of simply as:  Google is to Android is to your mobile phone as Microsoft is to Windows is to your PC.  While Google does sell Google branded hardware, it was quite happy to partner with other companies to do the work of manufacturing the phones.  (Samsung and HTC are prime examples of Android phone makers.)

There are dozens of makers of phones, but there are fewer makers of operating systems.  As far as Operating Systems go, currently Google Android and Apple iOS dominate the market.  The smaller players in this space are Microsoft (with Windows phone OS), Blackberry (with Blackberry phone) and I believe Nokia is still making a phone with its own operating system.  Obviously, Microsoft does not want to be a "small" player, and it's put a lot of energy into integrating its PC operating system with its Windows phone operating system (Windows 8 and Windows phone look and function almost identically).  Unfortunately for Microsoft, it's new operating systems don't seem to have attracted much excitement or enthusiasm.  With its mammoth presence in the PC market, PC users don't have much choice.  It might be good to recall that 2013 was the worst year in a long series of years of declines in PC sales.  I think Microsoft will never go away because it is the de facto choice for many business decision makers who want to avoid looking too snazzy, and it is the de facto choice for many average folks who want something "standard".

Long digression, now back to Google Voice Recognition….

Voice Recognition is integrated into the Operating System on all Android phones.  When I say integrated, I mean it is a button that is available to select on multiple applications.  For example, you can select it from within Messaging, Email, notepad type programs, etc.  It's built right into the keypad area and looks like a little microphone.  When the user taps the microphone icon, the application launches and after a pause where it says "getting ready," it asks the user to "speak now."  Once the spoken message is completed, the phone displays a "working" message.  After that, the text of the spoken message will appear.  Now, the text is not perfect but it is incredibly close to right, I would guesstimate 95%.  You just need to read over what was dictated and make a couple edits and your message is good to go.  Or you can just send it with that easily understood wrong word included.  (Everybody does that…  predictive text, Swype, and Voice Recognition have made for a lot of error-ful messages; but what the heck, we get our message across well enough.  It's probably no worse than static or interference in a spoken communication.)

Here's the extra-magical part (all of this stuff is magical, but this is where it gets pretty unbelievable).  When Google Voice says "getting ready" and when it says "working", it's communicating with a very very very powerful computing array called a "Neural Network."  Your phone is sending the voice recording to these computers which are then converting them into alphabetic words.  The neural network computers perform extremely sophisticated operations to decrypt your words, then re-send the alphabetic words back to your phone.  Now THAT is really cloud computing!  The transfer, the computing, the accuracy — all of it are marvelous.

So here is a link to a Wired magazine article about the neural network type computers, and prominently featuring Voice Recognition.

Final note from VHM:  The wondrousness of Google Translate, which never ceases to astonish me, must rely on the same sort of Cloud Computing that Tom has described for Google Voice.  Praise be to the Cloud!

Share:



41 Comments »

  1. Andy said,

    January 22, 2014 @ 3:39 pm

    While Google Translate uses the cloud by default, the Android app allows you to download large packs of data for each language and subsequently do translation entirely offline between those languages; no cloud required! I don't know if the quality of the translations are any worse if you forego the use of their powerful servers this way.

    Android also supports Chinese input by speaking, or even by simply drawing the characters (and selecting from a list of the closest matches if the phone's best guess is not correct). These are almost certainly slower or more error-prone than someone who is practiced with either pinyin input or your colleague's shape-based system, but have the advantage of requiring little to no training.

    I'm in my 20s, and technically inclined enough to understand the basic workings of all this stuff, but I still find it marvelous as well.

  2. Mark Young said,

    January 22, 2014 @ 3:44 pm

    so I have this on my phone and I can use it to type in a comment to pretty big

    not quite 95 percent — there were some drops at the end

    pretty close

  3. AntC said,

    January 22, 2014 @ 3:58 pm

    re Voice recongition:
    Is it just me, or is there something nuts about starting with a medium (voice) exquisitely adapted to convey all sorts of nuance and a device (a phone) precisely designed to convey that nuance; then ending up with a medium (text, or I should say txt) which far too often fails to convey that nuance?
    If you want async. communication, why not record a voice message on your phone, and send that to a voice inbox?
    In this day and age do we really have to worry about saving bandwidth (or cost) by using txt rather than voice?
    Call me Luddite …

  4. richardelguru said,

    January 22, 2014 @ 4:03 pm

    The voice recognition on my iPhone is pretty good too.
    I recently had it recognise 'haggis'! Of course it can also make some real schoolboy howlers.
    Now the real question for Language log is: Do we still call them cupertinos when it's from voice recognition?

  5. Giles said,

    January 22, 2014 @ 4:04 pm

    One Swype-like Android keyboard app that might be of interest from a linguistics viewpoint is Swiftkey.

    It has the same "drag your finger over the letters" system as Swype. But when choosing which words a particular pattern of dragging might represent, it not only weights its choices if by its own built-in estimates of probability — it also keeps a record of words you tend to use together and uses that to work out what you're likely to be saying. If you've just written "good" then pretty much any random squiggle starting on "m" and ending on "g" will get "morning". Indeed, it will probably suggest "morning" before you've even typed anything after "good". Additionally you can point it at a corpus of stuff that you've written (email archives, RSS feeds, etc) to train it.

    The net result is something that's often quite scarily accurate.

  6. Jim Breen said,

    January 22, 2014 @ 4:49 pm

    I had an HTC Wildfire too, which came with Swype preloaded. I tried it a bit, but as I don't type much on the phone, it was just a novelty. I knew about Swype, because I'd had discussions with the people there about adapting it to Japanese input.

    I was disappointed to find that when my Wildfire died and I replaced it with the HTC Desire C, Swype was no longer preloaded. Probably HTC has found it didn't offer enough value-adding.

  7. Pearl said,

    January 22, 2014 @ 5:05 pm

    I would like to add to what Giles has said regarding Swiftkey by mentioning that in addition to an existing corpus, Swiftkey now has the capability to use "Trending Phrases, based on current news and what’s hot on Twitter." (http://www.swiftkey.net/en/) So if there's a huge auction one week for Warhol's work, but two weeks later Andy Samberg has a new song with The Lonely Island that goes viral, the word predicted after you type "Andy" will likely be "Warhol" during that first week, and Samberg when that video gets popular.

  8. Ken Brown said,

    January 22, 2014 @ 5:14 pm

    I prefer text to voicemail. Not because it saves bandwidth but because it saves me time. I read faster than other people speak. Maybe ten or twenty times faster.

    Also the extra nuance of face-to-face conversation is mostly lost on phones due to poor sound quality and lack of non-verbal cues. And entirely lost in voicemail because most people use stilted language and miss out the important bits. Even me.

  9. Rubrick said,

    January 22, 2014 @ 5:16 pm

    I'm prepared to be corrected on this, but I'm pretty sure that the description of Google Voice offloading its processing to a "Neural Network" is incorrect. It does indeed do its processing via a massively-parallel, distributed computing system (as do most of Google's workings, one way or another), but that system is not configured as a neural network.

    Google does have a project in which it's using a subset of its vast computing resources as a neural net, called Google Brain; but I believe it's still in the pure-research phase, and is not being used in any shipping products.

  10. Keith Trnka said,

    January 22, 2014 @ 5:32 pm

    It's awesome to see Swype on LL – I work on the language models in Swype! If you've got questions, I'll answer what I can.

    @Giles Swype does that as well. Though maybe not as well known.

    @Rubrick Agree that calling the computing resource a neural network is weird. I've heard that they use deep belief networks (most ASR systems these days do), so it may be correct to say that it's using a neural network.

  11. AntC said,

    January 22, 2014 @ 6:40 pm

    Thank you Keith, and good to see you on LL.
    I was going to ask if there was a missing vowel in your family name (pssbly a prdctve txt flr ;-), but wikipedia tells me that there was even a Czech linguist Bohumil Trnka.

  12. Mark Mandel said,

    January 22, 2014 @ 6:44 pm

    I use both Swype and Dragon speech recognition (also from Nuance) on my Android phone. Being long accustomed to referring to speech recognition errors as "speakos" (by analogy with "typo"), I now also refer to "Swypos".

    I also use Honso's Multiling Keyboard, which claims to support over 130 languages (vs. ca. 50 for Swype).

  13. Keith M Ellis said,

    January 22, 2014 @ 8:43 pm

    "…it also keeps a record of words you tend to use together and uses that to work out what you're likely to be saying."

    If you aren't unnerved by the privacy implications, you can also give SwiftKey access to your gmail account, where it will use the corpus of your email for its word prediction. My experience is that this improved it quite dramatically.

    "I'm prepared to be corrected on this, but I'm pretty sure that the description of Google Voice offloading its processing to a 'Neural Network' is incorrect."

    I thought the same thing, but if you follow the link to the Wired story Victor provided, you'll see that Google describes the work they've done on Android voice recognition as utilizing a neural network architecture.

  14. Keith Trnka said,

    January 22, 2014 @ 8:53 pm

    @AntC hah, probably we're not related. But I don't know the family history before immigration to the US.

    @Mark Impressive language selection in Multiling. Many of those languages don't even have full font support in Android. It looks like they're a pure keyboard without correction/completion, which explains a bit. It's gets tougher and tougher to develop word lists, build corpora, etc. for additional languages.

  15. Dave Cragin said,

    January 22, 2014 @ 9:56 pm

    A book on translation called “Is that a fish in your ear?” has a discussion on how Google translate works. The author Bellos notes that Google uses software developed by IBM in the 1980s. Rather than try to translate a phrase itself, Google primarily searches the web to determine if a phrase has been translated before and then it mathematically picks the most common translation.

    It uses everything translated by the EU into 2 dozen languages since 1957 and everything the UN has translated into its 6 official languages. It also searches the web for translated works, i.e., books, articles, etc. “Drawing on the already established patterns of matches between these millions of paired documents, Google translate uses statistical methods to pick out the most probable acceptable version.”

    For translating between 2 languages that aren’t typically translated, say Icelandic to Farsi, Google uses “pivot” texts, such has English language novels that have been translated into both languages (a new use for Sherlock Holmes).

    I saw a wisp of how Google translate works in an obvious mistake it made. I was translating a simple Chinese sentence that included "Beijing and Shanghai" 北京和上海 but in giving the English, Google just gave "Beijing." I was dumbfounded, until I realized that a human translator had likely made this error translating the exact same sentence in a common text that Google found. I should have kept a copy of the full sentence, because it would have been good for illustration. If I had phrased the Chinese sentence slightly differently, I'd guess that Google would have “known” to translate 北京和上海 into Beijing and Shanghai (because it "knows" these words).

    (in the above, I'm summarizing 5 pages of text – the book gives more detail.)

  16. Dave Cragin said,

    January 22, 2014 @ 10:29 pm

    Here are a few more quotes on Google translate that may be of interest: "GT deals with translation on the basis not that every sentence is different, but that anything submitted to it has probably been said before. Whatever a language may be in principle, in practice it is used most commonly to say the same things over and over again." (Bellos, 2011). This makes me wonder about voice recognition, i.e., is it doing the same? – looking for patterns to ensure it recognizes speech accurately? (in addition to that mentioned in the Wired article).

  17. John said,

    January 22, 2014 @ 10:41 pm

    Predictive software has a ways to go, I think. I turn it off whenever I get a new phone, in fact. There are reasons…

    http://www.damnyouautocorrect.com/

  18. Mark Mandel said,

    January 22, 2014 @ 11:49 pm

    @Keith: No, Multiling has prediction. I don't know how the guy manages it in a free app, but he does. I'd use Multiling more than I do if I didn't keep getting the conventions and accesses mixed up with those of Swype.

    @Victor: It would help if you had some indication of the beginning of Tom's text, even just an <hr>. It took me quite a while to realize the change of author. It's at "Before we do anything else", isn't it?

  19. Mark Mandel said,

    January 22, 2014 @ 11:51 pm

    @John: Put not your faith in spellchuckers [sic], for they know not a word from its homophone.

  20. Matt_M said,

    January 23, 2014 @ 1:38 am

    @ Mark Mandel: "chuck" is a homophone of "check" now? What kind of accent do you have?

  21. Matt_M said,

    January 23, 2014 @ 4:03 am

    I've found that Google Translate is absolutely woeful at translating Thai text to English (my Thai friends assure me that the reverse is also true). It produces gibberish that, at best, might allow you to guess at the general topic of the original text, but nothing more. I wonder why it's so much more effective at translating other languages than Thai — is it that the corpus of Thai-English translations available to Google is tiny? Or is it something about the grammatical differences between Thai and English? Or the fact that the Thai writing system does not separate words by spaces?

  22. Ginger Yellow said,

    January 23, 2014 @ 5:27 am

    I was disappointed to find that when my Wildfire died and I replaced it with the HTC Desire C, Swype was no longer preloaded. Probably HTC has found it didn't offer enough value-adding.

    HTC's current keyboard software has a Swype-like tracing system, so they probably just decided not to pay the licensing fees.

    Personally I'm not a fan of the Swype system, at least on big phones. I find it much slower than tapping the keys (I'm a good touch typer on a PC, which may be part of the reason), especially with a good predictive system like Swiftkey. I can see the appeal on smaller phones where individual keys are smaller targets and they're closer together, I guess, but on a 4.7" screen there's no upside for me. It takes longer to push my thumb to the next key than to lift and tap. I also find I'm concentrating so hard on the pushing motion that I forget to look at the predictive suggestions, so I'm much more likely to have to type a full word out. That would change with practice, of course.

    Is it just me, or is there something nuts about starting with a medium (voice) exquisitely adapted to convey all sorts of nuance and a device (a phone) precisely designed to convey that nuance; then ending up with a medium (text, or I should say txt) which far too often fails to convey that nuance?

    Voice recognition on your phone isn't just (or even primarily) about communication with other humans. It's used for search, app launching, note taking, appointment setting etc.

  23. Nipo said,

    January 23, 2014 @ 6:57 am

    @Dave Cragin

    That explains something I've wondered for a while. When, during the summer 2011, I was moving and searched for "muuttoauto Helsinki Jyväskylä" in Google, I got "moving car chicago ny" as the translation. I couldn't understand why the city names were substituted, but a short search revealed there were several similar city name pairs where a city name was "translated" in the similar manner. I tried it now, but it seems that this has been corrected.

    Google is so very friendly and nice, trying to help the poor me by correcting my typos (aka foreign words) to legible and correct English words…

  24. Nipo said,

    January 23, 2014 @ 7:10 am

    Sorry for the repetition in the post above. My fingers left my brain behind againl

    About SMSing:
    As I use three different languages more or less daily, I find that changing the keyboards, languages and dictionaries for every message is more troublesome than just typing the text normally. Maybe this is why I never really learned swiping.

    I do have to confess that I love the idea behind TouchPal Keyboard (available at least for Android), which lets me access double the normal amount of symbols in the same keyboard. Tapping keys the old fashioned way lets me type words, and swiping the same keys downwards gives me numbers and punctuation. I really can't understand why the modern keyboards (save old Nokia ones, I think) need to hide periods and commas to other sheets.. no wonder the punctuation is getting out of fashion!

  25. Victor Mair said,

    January 23, 2014 @ 8:18 am

    @Mark Mandel

    "…indication of the beginning of Tom's text"

    Good suggestion. I didn't want to indent or italicize all of that.

    Fixed now.

  26. V said,

    January 23, 2014 @ 8:41 am

    How many of you have heard of http://en.wikipedia.org/wiki/Dasher_%28software%29 ?

  27. Stan Carey said,

    January 23, 2014 @ 10:15 am

    I proofread texts prepared using voice recognition software, and so come across some amusing errors from that wayward 5% (or thereabouts); the best recent example was "beaver coffee", which from context I inferred as "leave a copy".

  28. D. Sky Onosson said,

    January 23, 2014 @ 11:53 am

    I'd like to point out that the Google keyboard available for all android devices has a Swype-like input mode. I do prefer Swype myself over both Google's and Swiftkey's offerings, I'm using it to enter this post in fact. One of Swype's nicest features for me is a gesture (short swipe from the left into the spacebar) which toggles between the two most recently used languages.

  29. Dan Lufkin said,

    January 23, 2014 @ 12:14 pm

    The first predictive typing software I ran into was FlashForward, a plug-in for WordPerfect, which flourished briefly in the mid-80s. It seems to be gone now but something called WordLogic (q.G.) appears to be very similar. I used FlashForward for a while and liked it. I abandoned it when my keyboarding skills improved but I still have an FF swag mug that I keep my IrisPen in on my desk.

    My Samsung Galaxy phone converts incoming voice-mail to text, which you can read discreetly rather than listen to the message. I've used Dragon dictation software for over 10 years. Wouldn't be without it.

  30. Ginger Yellow said,

    January 23, 2014 @ 12:46 pm

    the best recent example was "beaver coffee", which from context I inferred as "leave a copy".

    Maybe it was flavoured with castoreum.

  31. Keith Trnka said,

    January 23, 2014 @ 1:18 pm

    @Mark It's very impressive that they have a dictionary for all those languages. Maybe he's using Wikipedia? I can't think of any other resource that has that many.

    @Nipo It may or may not help, but Swype has a bilingual mode that recognizes against two active languages. I think SwiftKey allows you to use three languages simultaneously. Recognition accuracy will be lower but it might be worth it if it saves you switching language so often.

  32. Dave Cragin said,

    January 23, 2014 @ 10:05 pm

    Matt_M That Google is bad between Chinese & Thai: I doubt it’s because Thai words are hard to determine or for grammar reasons because google translate is not trying read a sentence, it’s primarily looking for patterns that humans previously translated. Hence, I think your idea that corpus of Thai-English documents is limited seems more likely (and many may be poorly translated). For other languages, there are likely many official translations that google can use.

    Since it constantly trawls the web for translated documents, that it corrected itself with Nipo's example makes sense.

    The translation book points out that even in English, defining what a “word” is isn’t always clear. For example is “to take out” 1, 2 or 3 words? “Did you remember TO TAKE OUT the trash?” versus “I promised TO TAKE take my daughter OUT to see a film.” The author gives a "word" in Hungarian that he used in his daughter's wedding "e'desla'nyame'knak" which means "to my dear daughter's husband, in-laws, and friends."

  33. Victor Mair said,

    January 24, 2014 @ 10:25 am

    BTW, Tom knows a fair amount of Chinese. Here's a brief account.

  34. hanmeng said,

    January 25, 2014 @ 7:15 am

    Last Dec. on the Taipei metro I was looking over a young man's shoulder as he used handwriting to input Chinese. (I was only struck by it because I don't text and was curious about kids today.) It didn't look "frenzied" or like a "strenuous effort" to me, and as a matter of fact I seem to remember a few other young people inputting characters that way. I then assumed it was fairly popular.

  35. Victor Mair said,

    January 25, 2014 @ 8:27 am

    @hanmeng

    I've seen a "few" people do it too. Most of them were, shall we say, quite actively engaged with their mobile device. The professor from National Taiwan University whom I stood next to and watched do it for awhile was conspicuously and decidedly energetic in the way she input characters on her cell phone.

  36. Victor Mair said,

    January 25, 2014 @ 8:47 am

    One other thing I forgot to mention about Tom's language skills as a small child was his ability to use limited vocabulary to express a wide variety of ideas and things. Once, when he was three and a half years old and we were on a bus from Watertown to Harvard Square, we were sitting right behind a man with a conspicuously gleaming pate. I don't think that, before that time, Tom had ever seen a bald person, so of course he didn't know the word for baldness in Mandarin, and he barely knew any English at all. Nonetheless, he was intensely interested in this strange (to him) phenomenon of a man without a single hair on his head. So Tom pointed to the man's head (the same way he pointed to the calligraphy on the wall a couple of years earlier and said "ZI4" ["writing"] — as a little boy, Tom was curious about everything and pointed beautifully in the direction of anything that intrigued him), and proclaimed very clearly, "jīdàn bóbo 雞蛋伯伯" ("uncle egg"). That really cracked me up, but he often did that sort of thing.

    Adults can use the same technique when learning second languages. I remember when I took a test of spoken Japanese back around 1993, I surprised myself by scoring high when I thought that I really wasn't that good. After the scores were announced, I asked the examiner why my score was so high, and I still remember her reply, "You are especially good at expressing yourself with limited vocabulary." Well, I do manage to get along adequately in quite a few languages, but I certainly don't consider myself fluent in many of them.

  37. kimberlee said,

    January 25, 2014 @ 12:37 pm

    I'm sad to say that on my sg4 I was using the FREE version of swype since I had first purchased my sg4 last summer. Then all of a sudden 2 weeks ago it stopped working or unloaded itself off of my sg4 and I am VERY UPSET bcuz now I see swype is no longer FREE. I HATE my sg4 bcuz its always doing stupid things that I don't give it permission to do but it also continues to delete apps off of my sg4 that I was happily using. I just want my FREE FOREVER version of SWYPE back. Does anyone know how to get my FREE version of SWYPE back?
    Thank you in advance.

    PS… I always seem to get lemons everytime I get a new cell. I seem to have a black cloud. Even store agents are blown away with the weird stuff that my cells seem to do on their own. Its crazy but it really happens.

  38. kimberlee said,

    January 25, 2014 @ 12:47 pm

    PSS… Another thing I forgot to mention is not only did my FREE swype stop working and unload off of my sg4, now when people text me the time of their text shows up to from 30 to 60 minutes different in time. If the time on my phone says 9:37am and I've just replied to a text, when the person replies back my time will show 9:37 and the time on the text will show 10:37. I've taken several screen shots to show that there are 2 different times on my screen. Like I say, I always get lemons and our haunted cells that do what they want and such weird stuff that nobody has seen or heard of. Prior to swype disappearing my sg4 kept shutting itself off. Now that problem stopped only for this new problem of swype disappearing.
    Can you please email me with an answer of how to get my Free FULL version of SWYPE back.
    Thanks again in advance!
    Kimberlee :)

  39. Matthew Stuckwisch said,

    January 25, 2014 @ 12:56 pm

    For the people who have mentioned Honso's Multiling keyboard, be warned that it may be unusable for many of the languages it claims to support.

    I tried out the Asturian plugin and I'm not sure where it sourced its vocabulary. Since the programmer likely doesn't speak Asturian (a reasonably conclusion, methinks), the only source I can think that he'd have available is SoftAstur's original aspell dictionary. However, the Multiling keyboard doesn't handle any inflections or enclitics at all which makes it unusable. The original aspell one could at least handle either inflections or enclitics, but not both due to limitations in the aspell engine.

    (full disclosure: I'm working on an updated spell checker using hunspell that better handles the full breadth of Asturian's morphology and soon should have an Android version with keyboard out).

  40. Mark Mandel said,

    January 25, 2014 @ 3:41 pm

    @Matt_M: Oh, I distinguish /t͡ʃʌk/ and /t͡ʃɛk/, and always have. "Spellchucker" is my attempt at a derogatory pun, using "chuck" as 'throw away, discard'.

  41. Digitalis said,

    February 1, 2014 @ 2:13 pm

    I fee like I should just point out that:

    (1) Swype was NOT developed by Nuance, as Tom alleges. Nuance, owners of Dragon Naturally Speaking, purchased Swype several years ago (2011) from the creators–those creators include some of the team that first invented T9 input (we'll remember that from our old phones where we actually had to push the number keys to enter text, like on old Nokias). Swype has been around since at least 2008, and was first introduced on the Omnia in 2009.

    (2) You say something at the end about cloud computing and Google Voice, but I should warn you that Google Voice is NOT Google Voice Recognition. Voice Recognition, from Google, has been collecting data since at least the introduction of Google Voice. Google Voice is the company's voicemail and text messaging product, which has a feature that transcribes voicemails and allows you to mark those transcriptions for review and research. It is not equivalent to Voice Recognition, though they share data. They're using vast stores of voicemail data from Google Voice to improve their Recognition software. By comparison, Swype's nature voice recognition uses a different engine than Google's… Swype uses Dragon Dictation from Nuance, which has been around since at least 2000 (and probably earlier) in the software space for computers, with special emphasis on being an aide for the disabled.

RSS feed for comments on this post

Leave a Comment