- Our doctor makes housecalls
-
Reason # 6,384,289,121 to work at Microsoft: our healthcare includes free home visits by a doctor. We tried this today when my daughter had a scrape that seemed infected. One call to a special 24-hour number and an hour later a doctor was at our house. It almost felt like Little House on the Prairie, only instead of a horse-drawn carriage he came in an SUV. Yes, he did have a satchel with a whole bunch of medical items. After looking her over, he wrote a prescription and was on his way. For kids this is way nicer than an office visit, which can be pretty scary, especially if you go at a time when there are other patients around with serious conditions. And no need to worry about catching some other disease in the waiting room.
Speaking of healthcare, Bill Crounse, M.D. is Worldwide Health Director (and active blogger and podcaster) in the Health and Life Sciences group at Microsoft. He's also a fan of speech recognition, which could help with the dispatching of doctors. The physician who visited my house today coordinates with the other 3 doctors who are on-call today, deciding manually who will go to which patient next. It would be much more efficient if a Unified Communications server could keep track of locations and schedules, assigning new destinations via phone while the doctor is still on the road.
- Emerging Technology Conference
-
Looks like I'm going to the O'Reilly Emerging Technology Conference next week in San Diego. The theme of the conference is "magic", as in Arthur C. Clarke's famous saying about how sufficiently advanced technology is indistinguishable from it. We had hoped my boss and world-famous speech scientist Fil Alleva would be presenting, but they'll get me instead. What could I possibly say about speech as an emerging technology?
I'm thinking I'll talk about how speech is emerging as a bona fide data type, like numbers or text or dates. In SAT terms, I'll say:
speech is to voice as text is to fax
When you push somebody's voice through a phone line or you save it as an audio file, you end up with the kind of results you have when you send text through a fax machine: a pile of lifeless bits. Print an email message and fax it; what have you got? A piece of paper that can be read by one person at a time. Sure, you can modernize fax with color, fancy routing software, "paperless" delivery and more, but at the end of the day the content is just a bunch of bits. Same thing with spoken audio: go ahead and compress it to make it go further, use better microphones to raise the quality, strap on fancy new VOIP features to make it easier to pass it around networks, but it's still just a bunch of bits.
But imagine what happens when audio is transformed into its semantic content. To use the fax analogy, note how through email that same message can be sorted and searched, routed automatically through metadata like sender/receiver or by keywords, augmented with other data types like sound or video -- tons of features that make it far more useful than fax. Similarly with speech: what if I can transform an audio stream into its semantic content that can be searched, sorted, classified, and then transformed into other data types? It'll be like magic!
One nice thing about the conference is that it looks like my old Stanford roommate, Bill Katz is a speaker too. And Doug Cutting, who I worked with at CSLI for many enjoyable summers. Can't wait to see them!
- SR video from TechFest
-
Robert Scoble uploaded a PodTech video of his visit to Techfest earlier this month. Be sure to look at two sections:
- 34:04: Speech recognition for podcasts.
- 36:50: Frank Seide shows video exploration and discovery for Media Center PCs.
- What Tellme Employees think about iPhone
-
Hey, do Tellme employees agree with me about iPhone?
Here's a video posted to Youtube's tellmezone:
Like they say, "any phone can do that"
- ResponsePoint
-
Our old boss, XD Huang (yes, the author of the famous text on SR) has been hard at work the past few years on a new product, ResponsePoint, which was announced today. It's an IP phone designed for small businesses. Yes, it includes voice-activated features, but that's not the main reason you should buy one. It's VOIP (of course) and brain-dead simple for a small business to operate. Need to add or move an employee? Create a call-list distribution system? In the old days you would have needed a full-blown PBX or an expensive contract with a phone distributor. Now, you can just get ResponsePoint, from one of several OEM manufacturers including Quanta, D-Link, and Uniden.
Robert Brown (former colleague from our team who moved to XD's group last year) is blogging about it now.
- Array Microphones
-
Microsoft Research’s annual Techfest is the one of the busiest times of the year for my friends in MSR, who spend months building some the most amazing demos you can imagine. Some of our speech friends there, like Ivan Tashev and Mike Seltzer were featured in a Seattle P.I. newspaper article this week (check it out here).
Ivan knows more about microphones (esp. array microphones) than anybody on earth, and he came over this afternoon to present more to our team. Since 25% of new laptops these days come with array microphone support, he thinks there are lots of nice things we can do in communications and ASR to take advantage of that.
MSR people like Ivan are very approachable, and unlike some companies they are specifically encouraged to publish their results publicly. you can read a lot more from him on his web site or from the many whitepapers Microsoft publishes on How to Build and Use Microphone Arrays in Vista.
- Tellme Wow
-
It's official: the Microsoft Speech Deal of All Time: we announced that Tellme is now a part of Microsoft. My boss is in Mountain View today to meet the new people.
You'll hear a lot more about this in the next few weeks. And meanwhile, I'm not the official voice of Microsoft on this, so everything I say should be taken as my own personal opinion (and plenty of people are expressing their opinions).
But here's briefly what I think: Data is what drives speech recognition, and Tellme has it by the truckful. It's extremely important in speech to get actual users using your system, because every call and every interaction is new data that will improve quality and accuracy. A significant fraction of the U.S. population has called Tellme, which is an asset that will mean huge improvements to all of our systems.
I'm also looking forward to working with the people I know from my old days in Silicon Valley. Their offices are just down the street from my old neighborhood, so I get misty-eyed just thinking about Silicon Valley and all the fun things they do down there. Turns out there are several ex-Microsoft people there too, which should help in the transition because it means they already have a lot of the same DNA.
update: I keep mistakenly spelling 'Tellme' incorrectly with an inserted capital M, so I edited this post to use the correct orthography.
- What it feels like to merge with Microsoft
-
The last time I quit my company to look for a job, I joined a small start-up company in Silicon Valley. It had fewer than 100 employees and everyone was extremely smart, hard-working, and in love with our technology. Many of us had specifically chosen to work for a startup because it was a start-up: zero bureaucracy, common mission, small org where everyone knows everyone.
When a big "change in financial structure" happens, only a few key people are involved in the discussions (for obvious reasons). But I worked in the business development group at the time, so we kindda knew that something was up long before any official announcements. The company CEO was doing secret Sunday afternoon meetings (with Bill Gates, it turns out), mysterious guests would arrive from out of town and stay holed up all day in a conference room, and my VP boss was always flying back and forth from Seattle.
Once the deal was signed, a few of us were told the details so we could contact our customers and partners to give them a secret heads-up. Then the marketing, PR, HR, and others teams all went into overdrive to prepare for the public announcement. Of course the news was leaking quickly the whole time, so we all had to pretend nothing was up when the press called, while at the same time being sure not to say anything that was a complete lie.
Then one day the CEO called everyone into a mandatory company-wide meeting at a nearby hotel ballroom (our offices didn't have any conference rooms big enough) and he broke the big news. Then he had to leave early to join with various big shots at Microsoft to do a press conference.
The next few days and months were extremely fun. We were the center of attention in all the press, customers and partners -- who previously thought of themselves as the Big Guys and us as the Little Guys -- suddenly switched roles and were eager, almost desperate, to expand their relationships with us. But the best part was suddenly having access to all the wonderful resources of Microsoft, while still having that start-up soul.
That's why I still don't use the word "acquisition". A good relationship is more like a merger: they get something and we get something. And for me, at least, it's worked out pretty well -- I'm still here, ten years later.
- XNA game developer finalist uses speech
-
Heiko Rahmel pointed me to this new game by Afurtado as part of the Four-Day XNA challenge at the Games Developer Confernce. It uses our speech SDK.
- Speech Programming
-
I recently received the following question that Kevin submitted to this blog:
I'm interested in speech programming in c# and was wondering if you have some sample code for recognition in particular as i hope to develop a speech enabled ordering system where users could listen via tts and select items via recognition.
For some reason Kevin's email address didn't work when I tried to reply, so Kevin if you're still reading, please check out the following blogs for starters:
http://blogs.msdn.com/chuckop/
and the older one:
http://blogs.msdn.com/robertbrown/
which points to a great MSDN aticle
http://msdn.microsoft.com/msdnmag/issues/06/01/speechinWindowsVista/#void
As always, for desktop-related issues check Rob Chamber's blog and feel free to ping him -- he's very responsive.
Oh and one other secret email address: sapi5@microsoft.com Anything sent here will go directly to the speech developers at Microsoft, who are usually eager to help.
- Unified Communications at VoiceCon
-
Jeff Raikes, head of my division, just made a keynote address at VoiceCon Spring 2007, the big show for people in the voice industry. Since my group is part of the same division that brings you Office Communications Server, several of my friends are in Orlando now. Jeff says 100 million Office users will be able to make VOIP calls from Office in the next three years.
One new thing is that Jeff did a special Webcast in addition to the standard press release. Watch: in a few years that'll be a standard part of any press announcement.
- Microsoft Research Techfest keynote webcast today
-
Look at http://research.microsoft.com/ today at 9:00am (PST) to see Rick Rashid (head of Microsoft Research) kicking off our annual Techfest event. This is that time of year when MSR pulls together their coolest new demos and shows them off to (1) the world, through webcasts like this one, as well as press events, and (2) the various Microsoft product groups including mine, to stimulate us with ideas for exciting new products and features.
I'm planning to spend most of my Thursday at TechFest, but today's keynote will give you a good idea of what to expect.
- Summer Internships at Microsoft
-
Do you like languages? Do you like audio? Are you a year or two from graduating and starting to wonder where you'll end up? We're looking for students who want to work in my group for the Summer, and if you have an interest in Speech Recognition or Synthesis technology, please let me know.
Note: we'll bring people in from other countries too, so don't worry if your college is located outside the U.S.
- One more thing you can't do on Mac
-
David Pogue, the New York Times technology writer, gives another example of a cool thing available on Vista that you won't see on the Mac .
As always, Rob Chambers' blog is the best place to learn lots more.
- Ignite Seattle
-
While I was vacationing last week in Canada, I apparently missed one of the best Seattle events: IgniteSeattle. Fortunately, Lesley from our team was there and wrote up a nice summary of what I missed.
There were presentations about http://www.theworldisnotflat.com/ an interesting travel log. I read the entries about The Netherlands, which squares with my experience doing the MSTV office there years ago. I was there on business a lot, so I didn't see the libertarian side of the country, so my comment is that all of this may be true but it's invisible if you don't go looking for it.
TastingMenu.com has food reviews of many, many Seattle (and other city) restaurants. Yum, yum. Apparently the people who make that site did a presentation before Ignite Seattle -- wish I could have seen it. I wonder if it's posted somewhere.
Lesley knows about my interest in the Singularity and pointed me to the site http://www.embracingchaos.com by her friend Leo, who also presented something. One of these days I'll respond in more detail about my opinions.
The next event is on April 5th, so I'm going to mark my calendar and do my best to show up.
