Speech systems lawsuit sours Apple's day While the world glued itself to its iPad Mini press conference and complained that being an early adopter isn’t what it used to be, a Texas company filed a patent infringement complaint against Cupertino over Siri. Dynamic Advances alleges patents from the venerable Rensselaer Polytechnic Institute in Troy, New York – to which …
..........as its minimum claim."
So they are not claiming that they will be subjected to irreparable harm if they don't get well over a billion in damages then? Positively modest in fact. Perhaps they can get Judge Koh to hear the case.
The patent claims are very broad and on their own should be invalidated by decades of prior research. The concept of a natural language system necessitates a means of receiving the input and a database to match against. But the diagrams are fairly detailed, and if Siri does violate this patent, it is my uneducated opinion that Apple could make a work around.
It should be, but the same applies to most the patents Apple are trying their luck on. Anyone else I'd have been hoping common sense wins through, but Apple are hindering the industry so much nowadays I always get a (little) bit of kick when I see someone tries it on
Since the article says that not just Apple, but also HP, Microsoft, Sony and Google reference this patent in their own patents, apparently Apple isn't the only one who thinks this patent means something. If they beat Apple, seems likely they'd go after Google next.
Not sure what HP and Sony do with voice recognition, maybe some left over Palm stuff for HP and something on PS3 for Sony?
"Google's approach is much more functional, and faster"
I've played with VR software as a user off and on over the past twenty years, and nothing material has changed, other than that a mobile can now make the same daft mistakes that Dragon was able to make on a PC way back then. I don't see any worthwhile proportion of Apple customers actually using Siri, and fewer still of the Android user base.
Or to put it another way, they're both crap.
Apparently, in the US, there's not much you can't patent.
It used to be the states's way of granting an inventor a monopoly for a limited time, on genuinely innovative products.
Now it's a just corporation tool to hit other corporations, because they've been granted a patents for processes, event if there's no product.
"It used to be the states's way of granting an inventor a monopoly for a limited time, on genuinely innovative products."
And it can't even be used for that any more, because any large corporation will happily breach patents and then bury the Holder with legal costs if they want to get what they're entitled to.
That patent amounts to "receive a question, come up with one or more possible answers, then deliver one of them to the user", dressed up in legalese. You could probably just about demonstrate an Enigma machine as prior art ... indeed, arguably the Bletchley Park "bombe" decryption machines DO meet that, since the input is derived from natural language, they generated a finite number of possible outputs, then tried to identify one as the most likely decryption...
In fact, as faulty "patents" go, this one really is right up there with "rounded corners": whoever granted this should be fired.
That patent amounts to "receive a question, come up with one or more possible answers, then deliver one of them to the user", dressed up in legalese.
It does not. Have you read it?
There are a great many ways to implement natural-language query systems. While I won't claim the patent describes a novel one, it is quite specific about what sort of system it covers.
For example, claims 7 and 8 make it clear that their system does not rely on grammatical structure. That distinguishes it from the vast array of NLP approaches that begin by attempting to parse the input according to a model of the language's grammar.
Claims 9-13 show that their system generates a set of SQL queries (against the database of answers), then prunes that set to eliminate all but the most probably relevant. Again, that's one way to do it, but certainly not the only one.
Their system specifies learning behavior - it can add to the reference dictionary based on user input (claim 17). Not all natural-language query systems do.
Read the actual invention description, and you'll see that they're combining a keyword dictionary, prior-case information, metadata, a logical structure model, an interaction model, and a pruning algorithm (based on representing keyword/metadata/case hits as a graph and finding short paths). As they say, that's a good deal more complex than many mainstream NLQ systems of 2001 (when the patent application was filed), such as existing web search engines.
Their graph-reduction stuff might have been a novel approach to the problem in 2001, particularly when the graph is built from heterogeneous information sources. I really don't know. But it's certainly not as simple as far too many commentators have made it out to be.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
