Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
> ____/ Mark Kent on Sunday 30 September 2007 19:43 : \____
>
>> Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
>>> ____/ Mark Kent on Sunday 30 September 2007 12:04 : \____
>>>
>>>> Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
>>>>> ____/ Mark Kent on Saturday 29 September 2007 17:38 : \____
>>>>>
>>>>>> Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
>>>>>>> Nokia details Linux tablet WiMAX plans
>>>>>>>
>>>>>>> ,----[ Quote ]
>>>>>>>| The Nokia N-series tablets will integrate Skype, the Rhapsody music
>>>>>>>| service, and a Mozilla-based browser, Nokia has revealed. Mozilla's
>>>>>>>| Gecko rendering engine -- the same used in the popular Firefox browser
>>>>>>>| -- should deliver superior performance on Web 2.0 sites with lots of
>>>>>>>| AJAX, Nokia suggests. Apple's iPhone, meanwhile, uses a browser built
>>>>>>>| on KDE's lighter webkit rendering engine. Today, Nokia's N-series
>>>>>>>| tablets use a customized browser based on Opera's rendering engine.
>>>>>>> `----
>>>>>>>
>>>>>>> http://www.linuxdevices.com/news/NS4366436363.html
>>>>>>>
>>>>>>
>>>>>> And you can install minimo and w3m. Minimo is based on the mozilla
>>>>>> codebase, and works very well indeed. w3m is fine if you have a
>>>>>> bluetooth keyboard.
>>>>>
>>>>> The next tablet from Nokia (N800 successor) is said to have a built in
>>>>> QWERTY keyboard.
>>>>>
>>>>
>>>> Ah, so we finally get back to the clamshell of the Psion 5MX, then?
>>>> Hehe - so long it took!
>>>
>>> The keyboard will be needed until you can get a board powerful enough to
>>> cope with DragonNaturallySpeaking-quality speech recognition. IMHO.
>>>
>>
>> How will anyone be able to use such a machine in public, though?
>> Keyboards have the advantage of not needing any protocol to determine
>> from what source the information is coming, whereas voice-recognition
>> systems tend to assume that *all* information is aimed at it, and try
>> to determine what is noise, and interpret the rest as speech.
>>
>> If you consider this from a Shannon information theory perspective,
>> voice-recognition looks like an abysmal solution for anyone not in an
>> anechoic chamber.
>
> You could have a fallback, of course (or noise reduction). I've read a few of
> Shannon's original papers, but I'm not sure how this fits the problem at hand.
>
Information flows between terminating points. The entropy of the
informationn essentially describes how many different messages can be
sent. In the case of human communications, there is a vast range of
possibilities, although they are by no means infinite.
If you look at the communications protocols involved, then real people
will typically attempt to attract the attention of the person to whom they
are trying to speak, by perhaps calling a name, or using an honorific.
However, in all but the most formal of situations, much communications
is superfluous, aimed at a different listener or audience, or even just
plain wrong. Even meetings with formal chairmanship are not immune
to errors in this respect. Much human communications uses non-verbal
signalling of some kind to explain and clarify; for example, pointing,
waving, eyebrow movements (frowning, raising to show astonishment) as
we as mouth movements (smiling, being tight-lipped), plus head nodding
or shaking, arm movements, and even whole body movements.
So, the computer is on one end of a connection, but has virtually no way
of determining what is an "error" (ie., wrongly worded instruction,
say), or a misconnection (side-conversation with someone else, or
another person passing, or a train annoucement), a broadcast message, or
just some random noise.
Information theory would lead one to believe that the entropy of such a
channel is exceptionally high, and that the error rates and probabilities
of mis-connections are also exceptionally high. This would lead one to
consider that the actual throughput rate could be very low indeed.
Things get worse when you consider that almost all computers have
"modal" interfaces of some kind. It's always amused me that the
wintrolls will criticise Vi for its modal interface, and yet regard the
Windows GUI, which is also modal, as being fine.
If you're dictating to a word-proccessor, how does it check what you
said? What is the mode-changing command to switch to a clock or an
instant messenger say?
How do you protect against channel errors, misconnections, and so on?
If your "rm -rf /" command is made by accident just because of a
combination of words said by somebody behind you on a mobile, followed
by a train announcer, what will you do?
What if you're dictating to an email application, and then someone
behind you says "...escape... from here and ...send...", has your email
just gone?
How often are people irritated or upset because of some misheard or
badly put verbal communications?
What if the user is in a bad mood?
... and so on.
--
| Mark Kent -- mark at ellandroad dot demon dot co dot uk |
| Cola faq: http://www.faqs.org/faqs/linux/advocacy/faq-and-primer/ |
| Cola trolls: http://colatrolls.blogspot.com/ |
| My (new) blog: http://www.thereisnomagic.org |
|
|
