begin oe_protect.scr
Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
> __/ [ Mark Kent ] on Monday 31 July 2006 14:37 \__
>
>> begin oe_protect.scr
>> Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> espoused:
>>> __/ [ nessuno@xxxxxxxxxxxxxxxxxxx ] on Monday 31 July 2006 08:31 \__
>>>
>>>> I read an article in the newspaper recently about the great
>>>> improvements in voice recognition softwware for Windows, the bad demo
>>>> notwithstanding. Somewhat later I saw another article (probably on
>>>> cola) about great improvements in the same for Linux (free version). I
>>>> don't know how the two compare, however. Sorry I don't remember
>>>> details.
>>>
>>>
>>> Linux has some new speech recognition server. Being commercial, it should
>>> be far better than all the smaller projects.
>>
>> Not sure that I really subscribe to that assumption...
>>
>>>
>>> http://linux.sys-con.com/read/249714.htm
>>> <snip />
>
> Hadron argues that it re-uses IBM's famous work in that
> domain.
Having seen his attempts at reading and understanding networking, I'm
not sure that I'd really take anything he says very seriously.
> I can recall these tools from IBM going back to the
> days when I was 14 or 15. A friend of mine had this program,
> which occupied around 100 MB of disk space (quite steep at
> the time) and it requires _a lot_ of training. My cousin in
> Florida has been running Dragon NaturallySpeaking for
> several years, but he seems to only use it as computing
> vanity (i.e. when visitors arrive) rather than to get some
> actual work done, which dependent upon Dragon. And he's not
> that quick when it comes to typing, which pretty much tells
> you that this technology has a narrow/inexistent market
> (people remain curious nonetheless)... maybe a difference
> can come from miniature mobile devices -- those whose
> resources are not enough to make proper decisions and
> interpret voice correctly. Give it another decade maybe...
> assuming foldable keyboards don't see some other contender
> emerge. 3-4 years ago I had this silly idea about a PDA that
> binds itself to the wrist, so that you can input data with
> just a single hand (freeing the other for, e.g. umbrella or
> giving the finger to some troll on the other side of the
> street).
>
I think it'll get there in the end, but we're not quite there. This is
a good example of what I was describing previously, which is about
trying to make the correct decision on where the network-side processing
versus the device-side processing should go. For example, a device
might be powerful enough to stream the sound to a cluster which could do
recognition and message the result back.
This being the stuff that Mr Quark was struggling with.
--
| Mark Kent -- mark at ellandroad dot demon dot co dot uk |
You will overcome the attacks of jealous associates.
|
|
