Past EED rants


Live leaderboard

Poker leaderboard

Voice of EED

Wednesday, 2 February 2005

Need Input! [Slim]

The BBCs report on the health risk of the blackberry seems to conclude that the medium of email is potentially worse for your health than txt'ing. Kinda makes sense, more characters, more movement, more chance of RSI. Now there's been a lot written about viable alternatives to text input, and a bit about how new language is developing such as textspeak to make comms a bit easier, but really not much has actually happened to typing in the last few decades.
Take a blackberry, it's a remarkably small gadget that uses a querty keyboard, but that's not particularly innovative is it? Where's the fab new input mechanisms that we heard about? When will I be able to dictate into these things? Seems like we're making things smaller and smaller but not actually making them cleverer. Here's one idea for making txtspeak easier: why not translate it at the recieving end? Typing causes RSI, but reading doesn't. How abut a mobile device that turns 'u' into 'you' and spk into 'speak' when you're reading your messages?
Get your thinking caps on chaps, this tech's old and shit, we need new stuff if we're ever going to jet about in sky cars wearing silver track suits!


  1. The dumbest thing here is the idea of cramming all the keys of a qwerty keyboard onto a mobile device. People were quick enough to pick up predictive text (well, at least the non-stupid ones) so why not take a fresh look at the keyboard? These guys Frogpad have done exactly that. They make a little square bluetooth keyboard for one hand. Some claim they can type as fast on that as anyone can on a regular keyboard.

  2. You also get things called "chord keyboards", which basically have five slots for your fingers, and you apply pressure in combinations (like chords on a guitar or whatever) to create letters. Again, some people claim to have word-rates as fast as a regular keyboard.
    The Frogpad looks more interesting though. I'm quite keen to learn one of these actually - being able to type one-handed in that way would rock for mobile devices.

  3. I would certainly be a benefit to journalists who write consumer reviews for porn films.
    The frogpad is exactly what I'm talking about, someone needs to mass produce something like that shit. It's way too price as it stands to take off.
    I remember those chord keyboards on Tomorrows World years ago, they still around? They said at the time that they would revolutionise typing, but they never did. I guess it's tough to shift people off regular keyboards, which is why I recon putting a querty on a mobile gadget is such a wasted oppurtunity.

  4. Well, we never even managed to just alter the qwerty layout to something superior. With the Frogpad though, I thought that kind of thing is crying out as the solution for a portable messaging device.
    Anything's got to be better than trying to type on a Blackberry with pen caps taped to the end of your fingers.

  5. Someday, as proc speeds get better, the speech recognition people will get their heads screwed on right and (aswell as increasing accuracy) start releasing modules that can be plugged into apps on many platforms, and that use a definition file you can port from platform to platform once it's trained. (Platforms/input devices will probably have their own definitions too).
    You can train your voice on the PC, then "teach" your mobile device how to recognise your voice.
    This will happen - the sunrise happened too early for speech recognition, but it will make its comeback

  6. The answer is clearly mobile stenography.
    Or a mind reading thought-to-type helmet.

  7. Voice recognition is racist against northerners. I've not once found a package that will decipher my musical twang. I subject each one I try to the "apple" test. It usually ends up like this:
    Dave: "Apple"
    Software: "Ornament"
    Dave: "Apple"
    Software: "Quill"
    Dave: "Apple"
    Software: "Bridgehead"
    Dave: "Apple!"
    Software: "Smidgen"
    Dave: "Well fuck you then you piece of shit!"
    Software: "Apple"

    If thats the future, then I don't want to be part of it.

  8. It seems that the blackberry is going through a PR crisis.
    No less a tome than the Telegraph compares the addiction of regular blackberry users to crack addicts!
    Henceforth this most vein collapsing device shall be known as the crackberry.
    Word on the net though is that these surgeons are really just nokia fans and have only an interest in the crackberry so they can dis it. word.

  9. Heh - remember the first version of Dragon Dictate? fucking hilarious.
    A piece of software that proved once and for all that if you put monkeys in a room you don't get Shakespeare. You get Dragon Dictate, and it's about as much use a muslin condom.
    "NEW PAGE, NEW PARAGRAPH, BEGIN - The cat sat on the mat"
    "Sixteen pebbles lie inside the penguin's gullet".
    "Error. Command not found. Would you like a game of chess?"

  10. Problem with voice dictation is that it's a bit of a cock to use in public places. Dictating stuff on the train or whatever isn't really a solution. That said, there's some interesting work that's been done regarding sub-vocalisation - apparently it's entirely possible to pick up syllables from the throat when the speaker isn't actually pushing enough air through the voicebox to make an audible noise. You'd just stick a small filament microphone to the front of your throat and off you go.
    Would still look a bit odd because most people can't subvocalise without their lips moving, mind...

  11. That tech has been around the donkeys - it's what the military use. And Chuck Norris, in his excellent movies.

  12. I'd really like a Frogpad to use with my mobile, as using telnet from it is a bit of a hassle with the key layout of the Siemens SX-1. At work we've got a old Nokia Communicator 9000, and the qwerty keyboard on that works quite nicely, but then the phone is large as a brick...

  13. Voicerec is just horsepower innit and nuffin (really barring some decent dev time when commercial apps have unlimited horsepower at their disposal) well seriously nuffin else.
    The international security agencies Echelon has been chonking through several tens of millions of telephone conversations per month using the unbelievably huge combined budgets of those subscribing to it for several years now. At the parochial level, you get a clue about just how big the spend on this stuff is when you see that in the officially released budgets for the UK; GCHQ gets more money than MI5 and MI6 put together. Echelon, if you don't know (and no I can't be arsed to look through the web links to look for a decent one for it) is the pan-"western" nation global listening services that hoovers up billions of calls a year and can match both voice signatures and keywords to all of those calls which then feed into several layers of IT and then hum-int parsing. Participants are from the US thru Europe out to Aus and NZ. It's stations are deep satellite, line-of-sight satellite and then basic sit-on-top-of-the-incoming-cable-whacking-great-stations like those where the transatlantic cable comes into Southwestern england etc.....
    One of the most interesting recent publicised victims of Echelon was Kenneth Noye - serious villian, Brinks Matt robber (aquitted of the killing of a cop stabbed to death by him in his garden when raiding his house in persuit of the robbery) and convicted murderer on a bad day of someone he had a disagreement with on a M25 slipway. Noye is a serious Grade A gangster who commands big respect in whatever nick he's in and was always going to be protected when he did a bunk to Southern Iberia via a private plane when he was fingered for the crime. After several months of unproductive searching it was reported in the Times that his voiceprint had been caught by Echelon from a phonebox in Southern spain with the security services lending a hand to "settle old scores" for the Met. Having pinpointed the area of his calls, the sweep that followed netted him pretty easy.
    Echelon does mobiles, landlines and all the rest of it via brute force but also via tech which bears as much relation to the commercial world as a normal aeronautical engineer who could have claimed they could have designed an F117A twenty years ago. Indeed the work was real and possible but it took billions of dollars of tech to make it happen and was classified for over a decade and a half. Echelon's already "complete" and has a multi-billion dollar budget per year to keep "completing". You wonder how all those Al-Q cells keep getting caught? Here's a major clue.
    So whatever you do - don't say Osama set us up teh bomb....

  14. Thats all very interesting if you like reading Tom Clancy, but what does it have to do with redefining the keyboard?

  15. Or are you saying we should go put Memworth Hill in our rucksacks?

  16. Errr if you read the blog you'd seen we had a diversion into voice recognition stuff. It's one of the ways we entertain ourselves in our blogs. And we're able to hold more than one train of thought at the same time which is handy. And it's our webby so chuff off twathook...

  17. I think he means Menwith Hill... Concerning what Am said, voice printing is really really not the same thing as voice recognition. Voice printing basically takes a spectragraph, if you like, of the sound of a voice and compares that to another source. There's a service where you can hold up your mobile phone and record some music, where upon it tells you what the tune is. Same sort of thing. Voice recognition, on the other hand, is trying to actually make sense of speech. It's not actually hugely processor taxing these days, but it is one of the extreme applications of software engineering. It's not too bad to get it working, but it needs training. All the work so far is being on getting voice recognition so you don't need to train it to your voice.

  18. It's a fair distinction. What's interesting about the echelon stuff is that it not only does voice prints but in particular it does serious recognition *of a sort* in that it has thousands of useful definable words which it listens out for. So as well as matching prints relative to known recordings - Noye was picked up while using a public telephone box - it will hoove up all conversations where certain words are mentioned. That's from the obvious like 'bomb' to names of individuals and code words etc regardless of the speaker. Obviously the parsing that then follows before the millions and millions of conversations that are actively picked up each week are whittled down to a listenable quota is absolutely massive.