Apple acquired speech recognition firm Novauris last year to improve Siri - report
An Apple acquisition from 2013 may have quietly flown under the radar, as it was revealed this week that the Siri development team has added personnel from Novauris Technologies.
Apple's alleged purchase of Novauris was revealed on Thursday by TechCrunch, which reported that the acquisition took place last year but was not publicly announced. Personnel from Novauris are said to already be a part of Apple's Siri team.
While Apple has not confirmed the purchase, reporter Sarah Perez placed a call placed to the Novauris offices in the U.K., and an employee there confirmed that the operation is now a part of Apple.
Novauris was founded by the same people who founded Dragon Systems, creators of Dragon NaturallySpeaking, which is now owned by Nuance Communications. Nuance also specializes in speech recognition, and its technology helps to power Apple's Siri, the virtual personal assistant found on its iOS devices.
Nouvaris specializes in automatic speech recognition technology utilized to access information stored either locally on a device, or on remote servers. The company's website remains operational, and makes no mention of an acquisition by Apple.
Its featured product is NovaSystem, a scalable, server-based speech recognition system capable of handling multiple simultaneous voice access requests. The system can also spread the computational load over a multi-computer installation if necessary.
"Novauris believes that voice will become the interface of choice for consumer applications of all types," the website reads. "Users should be able to ask for information quickly, easily and directly without needing an enrollment process or complicated and time-consuming dialogues."
Apple spent $525 million on acquisitions in the last quarter of 2013 alone. Those purchases are believed to include personal assistant app Cue, 3D motion sensor firm PrimeSense, and social media analytics company Topsy.
Other 2013 purchases by Apple include digital mapping company BroadMap, public transit tracker HopStop, mapping data startup Locationary, and indoor GPS company WifiSLAM.
Apple's alleged purchase of Novauris was revealed on Thursday by TechCrunch, which reported that the acquisition took place last year but was not publicly announced. Personnel from Novauris are said to already be a part of Apple's Siri team.
While Apple has not confirmed the purchase, reporter Sarah Perez placed a call placed to the Novauris offices in the U.K., and an employee there confirmed that the operation is now a part of Apple.
Novauris was founded by the same people who founded Dragon Systems, creators of Dragon NaturallySpeaking, which is now owned by Nuance Communications. Nuance also specializes in speech recognition, and its technology helps to power Apple's Siri, the virtual personal assistant found on its iOS devices.
Nouvaris specializes in automatic speech recognition technology utilized to access information stored either locally on a device, or on remote servers. The company's website remains operational, and makes no mention of an acquisition by Apple.
Its featured product is NovaSystem, a scalable, server-based speech recognition system capable of handling multiple simultaneous voice access requests. The system can also spread the computational load over a multi-computer installation if necessary.
"Novauris believes that voice will become the interface of choice for consumer applications of all types," the website reads. "Users should be able to ask for information quickly, easily and directly without needing an enrollment process or complicated and time-consuming dialogues."
Apple spent $525 million on acquisitions in the last quarter of 2013 alone. Those purchases are believed to include personal assistant app Cue, 3D motion sensor firm PrimeSense, and social media analytics company Topsy.
Other 2013 purchases by Apple include digital mapping company BroadMap, public transit tracker HopStop, mapping data startup Locationary, and indoor GPS company WifiSLAM.
Comments
This is exactly what I hoped Apple would do, when I suggested it buy the Baker's team in my comments to an Appleinsider story last July:
http://forums.appleinsider.com/t/158729/apple-assembling-speech-recognition-tech-team-in-boston-to-improve-siri#post_2368625
...and in another of my comments last September:
http://forums.appleinsider.com/t/159303/investors-speculate-carl-icahn-could-push-apple-to-buy-nuance-for-speech-tech#post_2388324
The article triggered a memory of someone's purchase suggestion. I was going to look it up, but thankfully you made yourself known.
Thanks for the suggestion!
It's a bad thing that Siri doesn't work without a network.
Speech recognition should be done on the device and not on a server, that's also the best load balancing imaginable. A device with 115 GFLOPS should be more than capable.
It's a bad thing that Siri doesn't work without a network.
There isn't just a single way.
I mean, by this time, the majority of Apple's devices that are sold are connected to the Internet one way or another. I get it, networks suck sometimes, other times there's slow or no WiFi, but more often than not, if you're using one of these devices you're online.
You're also forgetting all of the benefits that come from the server communication. Apple can make updates to the Siri engine without ever needing to update the device, you can have machines dedicated and optimized for the processing of Siri requests (much more efficiently than the phones hardware would without added complexity, expense, etc), you can utilize the incoming data to get analytics, the phone doesn't need to store the massive amount of natural language-decoding software components... the list goes on and on.
There are obviously many other benefits and it's ridiculous to think Apple did not weigh them in their development decisions with Siri.
Speech recognition can be solved without a server, and current hardware can do it as good as any server.
Think of the network traffic and server capacity needed, it's all a waste because the computing power of all devices combined exceeds that of any server by far.
Bitcoin mining is now done on mobile phones, why is that you think?
Hard Proof that Apple executives read this blog to know what to do next!!! /s
Definitely getting more accurate and seems faster.
Anyone who thinks Siri is improved please try this test.
Me: "Siri, who is the president of the the United States?"
Siri: "Let me think about that...The answer is Barack Hussain Obama II" (she said Obama 'two')
Me: "What is the name of his dog?"
Siri: "Checking my sources... My web search turned this up."
Displays a link to Dog - Wikipedia.
But if you ask her "What is the name of President Obama's dog?" She gets it right, sort of right, "Bo" but she doesn't speak it just displays on screen. Actually "Bo" is only one of the Presidents dogs. "Sunny" is the other.
My conclusion: Siri still cannot carry on a conversation worth a damn when it involves retrieving information from the Internet..
Your results may vary, but I try this type of contextual query with Siri all the time and she almost never has a clue what we are discussing after the first question.
When you try the same test with Google Search the results are completely different and the there is no "Hmm, let ne think about that" either. The speed difference is like a hundred times faster too. Give it a try and then tell me which works better....I mean, which one works.
What Novauris does that Nuance does not...
Has anyone tried the 'Tell me a story' thing on Siri? It is pretty funny. You need to ask her thrice...
Speech recognition should be done on the device and not on a server, that's also the best load balancing imaginable. A device with 115 GFLOPS should be more than capable.
It's a bad thing that Siri doesn't work without a network.
I agree. The main reason I don't use Siri very often is I don't want Apple having a library of MP3s of my voice.
Has privacy been a concern? No. Maybe for the select few, but not the millions of others.
Can iOS be updated easily? It depends on what your definition of easy is. You don't even know how often they do make changes to the Siri engine - it could be everyday, it could be once as month, but I guarantee you it is much easier to update a server farm than change iOS and then push that update out to millions of iDevices frequently.
And no, iPhone hardware in it's current form cannot do Siri as well as a server optimized specifically for that task. By contracting out, your iPhone gets better battery life, isn't being stuffed with potentially unused software components, not using hardware to handle requests making the phone cheaper, etc etc. You also state that all counting power of these devices COMBINED is better than a server? That's cool, but you're also implying these devices to be independent processing requests. When multiple servers are pumping away at your single request, you can be damn sure they'll push out that request faster than locally on your phone, or at least fast enough that you'd often not notice network fallacies, which is the biggest drawback in this method.
Bitmining on phones is NOT cost effective or efficient at all. It's hardly efficient on most computers and requires specialized hardware to make it that way, which costs extra money, resources, time, etc. The same thing can be said for Siri on iDevice hardware.
Hard Proof that Apple executives read this blog to know what to do next!!! /s
Heh. I'm sure Apple execs have better things to do than read our oft-rambling comments. But doesn't it always feel good to see something happen which you felt ought to happen? That's all I meant to express.
I wonder if, at least part of, this acquisition isn't a subtle FU to Carl Ichan, Apple gadfly and large Nuance shareholder. I'm sure Tim Cook said "why buy the cow, when we can get Novauris for [comparatively] free?" The fact that it kinda screws a major pain in his ass may have just been gravy.
You are both clueless. Siri is not just speech recognition for one thing. It's an AI app that requires the server back end in order to aggregate and respond with the information requested with some level of intelligence based on previous requests, location, time, and other factors. The more it's used the more likely the response will be what your looking for.
It's unlikely that Apple has any interest in storing your or anyone else's voices in MP3 format from your Siri queries. I'm sure you used MP3 to make a point but it wouldn't be the chosen codec for storage due to space. And if they did save those hundreds of millions of voice queries, what purpose would it be for? A massive library of all of us saying "Siri, tell me a joke." when we first got our iPhone 4s and above. If you're afraid to talk to your phone to ask for information or directions then what do you think your ISP or cell carrier is collecting about you when you access the internet? Any app that uses location services knows more about you and where you are than your mom did the entire time you were in high school. Those ads you see aren't there by accident either.
You are both clueless. Siri is not just speech recognition for one thing. It's an AI app that requires the server back end in order to aggregate and respond with the information requested with some level of intelligence based on previous requests, location, time, and other factors. The more it's used the more likely the response will be what your looking for.
It's unlikely that Apple has any interest in storing your or anyone else's voices in MP3 format from your Siri queries. I'm sure you used MP3 to make a point but it wouldn't be the chosen codec for storage due to space. And if they did save those hundreds of millions of voice queries, what purpose would it be for? A massive library of all of us saying "Siri, tell me a joke." when we first got our iPhone 4s and above. If you're afraid to talk to your phone to ask for information or directions then what do you think your ISP or cell carrier is collecting about you when you access the internet? Any app that uses location services knows more about you and where you are than your mom did the entire time you were in high school. Those ads you see aren't there by accident either.
Yes, I just used the term "MP3" to make a point, what the actual codec is is irrelevant. What you perhaps don't know that Apple do in fact keep recordings of all queries sent to Siri for 2 years. This is from their own spokesperson, not a rumour.
http://www.wired.com/2013/04/siri-two-years/
You may choose to turn off Siri at any time. To do so, open Settings, tap General, tap Siri, and slide the Siri switch to “off”. If you turn off Siri, Apple will delete your User Data, as well as your recent voice input data. Older voice input data that has been disassociated from you may be retained for a period of time to generally improve Siri, Dictation and dictation functionality in other Apple products and services. This voice input data may include audio files and transcripts of what you said and related diagnostic data, such as hardware and operating system specifications and performance statistics.
The article reiterates that with a clarification of the time (2 years) and throws a bit of paranoia in as well.
If you are so concerned about Apple have clips of your voice (they have more and Google knows you better than you do) then just turn it off. Doing voice recognition is possible standalone but doing an application like Siri requires a host server right now and for the near future. If you're concerned about your private information being collected on a server somewhere turm off location services, don't buy any apps music or movies, don't use maps or Google maps, stay away from unknown hotspots that require a login or confirmation to connect and don't use any social media. Also stay away from any cloud services. Of course your cell carrier is going to collect information about you no matter what you do and chances are that your carrier, Google, Yahoo, Microsoft, Apple or Facebook has already handed your info over to the NSA. Personally I'm getting too old to worry about all this. I use Siri all the time. She gets more accurate as we get to know each other. I tell her how to be smarter by teaching her things like how to pronounce names and telling her relationships. It's much more fun when you relax and get to know her. When I got my iphone4s I would say "call Jill Johnson Mobile" to call my wife. Now I say "Get ahold of my Sweet Pea in the car please". I like to be pleasant with Siri. Or to text my daughter "Tell the kiddo I'm on my way and be there in 5". Sometimes I just ask for useless info. I learned this one the other day "what planes are flying over me right now". Just for fun.