|
Authored by: Anonymous on Wednesday, July 11 2012 @ 10:54 AM EDT |
Go back and read these patents. They don't claim, say, a
particular form of hidden Markov model combined with a
particular training dataset - those are some key, inventive
parts of a good speech recognition system. As a matter of
fact, those basic elements were published academically a
long time ago. If you tried to patent them, the patent
would be rejected. It's only recently that hardware and
networking is catching up to make those systems practical on
PCs and portable devices.
(A few years ago I worked on a prototype portable device
for the leading speech company at the time; that particular
prototype was never finished because the company went broke
due to criminal mismanagement, but it was intended to do
roughly what Siri does, except it was an offline system -
WiFi didn't exist back then.)
Patents like the one at issue here don't claim any of the
hard stuff, they just claim "a system that does X." They
list the input and the output, and that's basically it.
There's no invention in there at all. [ Reply to This | Parent | # ]
|
|
Authored by: jonathon on Wednesday, July 11 2012 @ 02:16 PM EDT |
>A method that allows a computer to recognizes speech and type what is spoken
didn't become obvious the instant a computer was invented.
The specific mathematical algorithm might not have been obvious, but since
mathematical algorithms are not patenable..
And since doing something "with a computer" is not inherently
patentable....
[ Reply to This | Parent | # ]
|
|
|
|
|