|Google registers voice-activated search patent|
|A newly published patent provides further evidence that Google is developing a voice-activated search engine. |
Patent number 7027987, of which Google is the assignee, concerns "a voice interface for search engines. Through the use of a language model, phonetic dictionary and acoustic models, a server generates an n-best hypothesis list or word graph."
Though the patent was only published this week, the application was originally filed in February 2001, indicating that Google has had the project in the works for some time.
Google is, apparently, suggesting that not too much should be read into this patent being published.
Very very tasty - they will be very pleased to have got that one in the bag.
While the World might not be ready for it quite yet, it's a no-brainer voice search will happen at some point. The possibilities of what you can do with it are boundless.
Good Luck Google!
This does not mean Google are working on voice search at the moment.
What it does mean is that Google realise that searching via voice will be technologically feasable within the next 20 years.
Google have 20 years grace which prevents any competitor from entering this market in the USA within the timeframe that the patent is valid for.
You dial a number. Or if you have a phone that's at not more than 10 years old, you can say a command.
Your phone dials Google.
Google answers. They ask you to speak your search words.
Their computer recognizes what you say.
It uses what you said to do what it always does - look for stuff.
It returns results, via existing methods, to your computer, to your phone, to the chip in your brain, or whatever else you can think of.
I don't see where the new technology is in this. This is another example of a patent that should be denied. Voice recognition by computers has been around for a long time. The practice of using voice commands to drive computer processes has been around a long time. And the act and art of tying this into an automated phone system has been around for a long time.
If their patent is on better voice recognition, then they should say so. If the patent is on a better search algorithm, then say so. If it is on localizing the search based upon the physical location of the caller, then say so. If the patent is on the transport process, then say so. But the patent is not for any of these things. The process (It is a process that is being patented) covered by this patent application is widely employed and has in fact been in use for quite a while.
Its a "one-click buy" patent for phone searches.
The heart of the patent is that they will be very thorough, and very good at determining what the person says, no matter what the language.
Their description of how this is done boils down to "we will apply very advanced methods and algorithms to figure it all out."
They don't actually introduce anything new.
It makes me seriously wonder how much stock the Primary Patent Examiner has in Google, Inc.
This is rediculous. So annoying to be able to file patent for something, admit that you aren't even working on it and "promise" to do a good job.
I know exactly how usable this feature could be and I personally have the resources to do equally as good of a job with it...and right now. Not 20 years from now.
And as the poster said above...you call in..speak your searc etc... You can do it via voip or pstn. The additional possibilities of sending the search results directly to your phone or your computer or converting them text-to-speech is actually endless. In that respect google does have infinite possibilities because they are spread so diversely.
|I personally have the resources to do equally as good of a job with it...and right now. |
You should have filed a patent. :)
IT appears from the patent that the main feature that does, ("shall" or "may"), differentiate the "process" from the simple joining of current voice recognition software as an interface to currently deployed searchengines is a weighting algorythm where the logic part of the process will determine the best match when you speak a few words.
ie- if your query is "washington clothes", some voice software may interpret this as;
- washing ton close
- washington close
- washing ton (of) clothes
The patent implies that a lookup will be done to see which makes the most sense, (they could employ a list of common searches from their logs, take into account date/time, ie- if the search was done near "George Washington's Day" (US), they could deduce it was a search for "information about period costumes from the US revolutionary era worn by George Washington".
If the search was done just after a search for "landry", they could assume you want to "wash a ton of clothes" and return a local search for a laundrymat...
If they get it right, it could help with regular phone / information calls --- I will admit to yelling at computers that answer the 411 calls ---- "No I said Mark Wright not Market Basket you stupid computer, put a human on the phone!"
Understanding natural languages has always been the Vietnam or Iraque of computer sciences.
Does anyone in here actually USE any of these existing voice-based tools yet?
I even got fed up with my thesaurus a few years ago. I used so many words in my written language, which it didn't know, that inserting these into the dictionary would have taken much more time than that saved by the programm correcting minor spelling mistakes. I turned it off immediately, and after these experiences on the mere lexical level, I decided not to buy any of these voice-recognition-based little helpers. Lexipixel's example gives a good glimpse.
Does anyone recall "You just don't understand." (1990) from Deborah Tannen? If google adds another patent capable of deciding male and female from who is calling, please wake me up.
unfortunately, you're mistaken. this isn't the iraq of computer science. this is something that's been studied for the past ten years--and there is plenty of development to move forward with implementation. think about everything that ibm, intel, and microsoft have done...
And, yes, I have used this resource. In fact, there are people in a cafe I own which use this on a daily basis--and thousands of others who do so as well. One of the significant aspects of this, is that it can help those who are impaired and be integrated further into existing electrionic functions (imagine talking while driving to the tune of google search).
I've seen this technolgy at work and, quite frankly, the fact that google has chosen to move forward with the exploration is a significant development.
on a somewhat related note, one of the ultimate gold mines is looking at google Jobs page. This is news I knew out about 8 to 10 months ago--given the type of job descriptions available. Not trying to brag but rather would point out that a lot of trends which are seamingly news worthy can be determined by an analysis of their HR pursuals. For their big endeavours, they occasionally resort to other firms for headhunting which gets even more interesting..
This will for sure be a part of Google's Mobile strategy. Voice activated search on your computer makes no sense.. there is little need for it to realize any sort of material benefit (who really cares) But to be anywhere with your mobile phone and enter for example, "#G" on your keypad and then say what you are searching for, and Google will then send you back a text message with a direct link to search results (wap push a link)... this could position Google as a global 411 directory... of course where you are calling from will factor what set of results are sent back to your phone.
I do appreciate what these programms might do for disabled people. I also think that there will be a number of forthcoming interesting applications concerning those voice-recognition- and location-based tools. Indeed googles announcements on tools for mobile phones should be followed with much care.
On the other hand I doubt anyone is using such tools for doing e.g. Php-scripting or other programming.You don't need to study linguistics to know that phonetic ambiguities will mulitply the sAND-problem and related issues. I guess for each problem solved (like the paradigm lexipixel gives) there will be two new ones you didn't have without voice-recognition. Natural language is a Hydra, just like Al Qaida seems to be.
Ten years ago: that's quite exactly the time when I stopped further research on the promises of understanding natural languages. Sometimes I'm quite fascinated when I see the progress, that has been made since then, for instance googles translation tool. But then I recognize all these funny little mistakes...
There are very fundamental limits to computers understanding natural languages. You can use ordinary language to discuss about language and build theories about language and in most cases it is implicitly clear what level you refer to in each sentence. Comparably, you can edit a text via voice in your preferred editor and additionally you have an escape-sequence for your format-options. Now imagine you were writing a manual describing how to use this escape-option. Try to do that without using your keyboard.
Would be interresting to try on practice