Look Out, Nuance: Apple's Office Near MIT Is Stocking Up With Speech-Tech Talent 50
An anonymous reader writes "Apple's had a small, very secretive office in Cambridge, MA for a few months now. And we finally know what they're doing: Building a team that works on speech technology for Siri. Sure, it's interesting for Apple to have a remote engineering team. And hiring from MIT is a no-brainer. But here's why this is a bigger deal: Apple has always relied on Nuance, a Boston-area company, for the speech-recognition technology behind Siri. By branching out with its own speech team — stocked with former Nuance scientists, no less — Apple could very well be signaling a move away from relying on Nuance for this core technology. And the speech wars are just heating up: Microsoft and Amazon both have speech engineering offices in the Boston area too."
Good move, Apple. (Score:1)
If Nuance was as good for Siri as it was on Android, an offshored data entry clerk who doesn't speak English is probably a better choice.
It added a special sort of pain when G took out the voice-dial confirmation prompt in Gingerbread (I think, maybe it was Froyo?).
Still won't buy an iPhone, but this is one case where I can't hold Apple's NIH attitude against them.
Re: (Score:3)
They need help with Siri (Score:5, Funny)
Voice wecognition on that thing is terrible. Wook.
Siwi, can you wecommend a westauwant?
I'm sorry, Bawwy. I don't understand "wecommend a westauwant."
Wisten to me. Not "westauwant," *westauwant*.
I don't know what you mean by "not westauwant, westauwant."
See? Total cwap. You suck, Siwi.
Re:They need help with Siri (Score:5, Funny)
And here's my favorite [amazonaws.com].
"Siri, I'm bleeding really bad, can you call me an ambulance?"
"From now on, I'll call you 'An Ambulance'. OK?"
(This was apparently changed in one of the updates.)
This may be speech recognition, but it isn't any sort of content recognition. It's just pattern matching, and only those patterns which the coders have anticipated.
Re: (Score:2)
Siri:"...Turn left in 100 meters."
Boston Dialect (Score:2)
Perfect Boston Accent [google.com]
Re: (Score:2)
Well, a better reason might be that nobody in North America calls it a "car park", and if you've got Siri set to US/Canadian, you're not using the right terminology.
Re: (Score:2)
We have car parks for city people who insist on walking around with those ridiculous toy cars, leaking all over the sidewalks and making all sorts of annoying noise. Of course, I think it is silly to use scarce open space in our parks for such frivolity, but enough people insist on having these little trophies that I'm apparently in the minority.
Re: (Score:1)
OK, I'll bite. Most of us call WHAT car parks? Are you talking about parking lots? Or places where VW Beetles can frolic and sniff the exhausts of bigger models?
Re: (Score:2)
No. We say we parked our car but we do that in a parking garage, parking lot or parking space.
Some houses do have carports though.
Re: (Score:2)
Europeans call them car parks. North Americans call them parking lots/garages/spaces. I don't think I've ever heard an American call anything a "car park" in casual conversation.
All speech recognition from Boston... (Score:4, Funny)
yeah, that's gonna work out well.
Re: (Score:2)
Last time I was in bean-town the airport rental car shuttle driver warned me about 'cops in tourists on the shoulder'. All I could say was 'I'm here on business. When did they make sodomy mandatory for tourists?' He was trying to say 'Taurus'' (which tells you long it's been sense I've been there).
Speech engineering... in *Boston*?! (Score:5, Funny)
And the speech wars are just heating up: Microsoft and Amazon both have speech engineering offices in the Boston area too.
"Siri, wheah's a wicked good place to pahk neah the Gahden?"
iOS for cars (Score:2)
It was on their last conference call that they believe it is very important. That's going to involve a lot of voice interaction.
Google Voice as a dev pool (Score:2)
I often wonder if Google Voice's transcription service for voicemail is a way for Google to get people to provide them with voice-rec feedback. They have those buttons to allow Google to use individual voicemail messages and transcripts to "improve" their service. You can bet they've got an angle.
Apple Maps mark 2? (Score:1)
Going it along didn't work out too well with their mapping software... I think they underestimated the difficulty of doing it well, and probably have done so again with speech recognition.
Re: (Score:2)
Re:Apple Maps mark 2? (Score:4, Insightful)
Re: (Score:2)
Apple has done in-house voice recognition in the past, although that was twenty years ago.
Re: (Score:2)
California?? (Score:2)
Re: (Score:2)
Re: (Score:3)
The speech work is being done in Boston because - they figure if Siri can correctly interpret words spoken by a Bostonian, it'll have no trouble with folks who speak actual English.
Re: (Score:2)
Actually, I worked for a company that did speech recognition as part of its produce many years ago. The found the hardest accent was Indian.
Re: (Score:2)
product, I meant, not produce!
Re: (Score:2)
And "apple speech recognition" takes on a new meaning.
Re: (Score:2)
Apple's peach recognition.
Re: (Score:2)
Re: (Score:2)
As opposed to Korea, not Boston.
these boston accent jokes (Score:2)
...would be funny if any of these positions actually went to native Bostonians. But they won't; no doubt they'll be mostly MIT/Harvard/BU students and alumni from all over the world. And none of them will be older than 30 years.
Not necessarily about MIT... (Score:3)
The real powerhouse in speech recognition tech isn't MIT -- it's BBN, at the other end of Cambridge.
Re: (Score:2)
Nuance isn't competition (Score:1)
Re: (Score:2)
Nuance bought the speech technology from Lernout & Hauspie when the latter went bankrupt after a fraud scandal. I don't have the feeling that they developed the technology a lot further it seems that it still is the "same" as a couple of years ago.
Doesn't surprise me as they bought the software, not the talent behind the development.
Cautionary tale of Bakers losing everything (Score:2)
http://www.bloomberg.com/news/2012-12-10/dragon-systems-founders-take-goldman-to-trial-over-advice.html [bloomberg.com]
"In a federal trial that began yesterday in Boston, the Bakers claim that shoddy work by Goldman Sachs on the $580 million all-stock sale of Dragon to a Belgian competitor, Lernout & Hauspie Speech Products NV, cost them their company and their fortune. Within months of the sale's June 2000 close, Lernout & Hauspie collapsed in an accounting scandal and its shares that the Bakers took as payment fo
Deconstruction algorithms. (Score:2)
I wonder is they are going the wrong way about it. The dialog's between the user's and Siri are
truncated and one-sided.
What Siri needs to do is to collect conversations for one or two years of millions of users and learn
to carry it's own conversations before is ready to help anyone.
Perhap's Siri needs some help from the people in the NSA.
Hmmmmm (Score:2)
Here's the question: are these people working on technologies to convert speech-to-text, or are they working on the next layer after that: parsing/understanding that text in a way to produce useful results. Given the state of the patent system, and the amount of IP Nuance owns, I'd be hesitant to even try to outcompete Nuance on the speech-to-text part. But there's still lots of work to do on what to do with the data that gets spit out by the speech-to-text processor. On the other hand, it's possible Apple