the spkydog koop

Microsoft to enter the smartphone business?

2009-02-07T03:57:00.001-08:00

Dell, now Microsoft? It seems highly unlikely, but there is some chatter out there.

Palm Pre is looking good

2009-01-13T08:26:00.000-08:00

The Palm Pre is a beautiful thing indeed. Even Walt Mossberg was impressed.

Take a look at phonescoop's review on youtube:

Then, check out the galleries on Palm's site.

Luxurious Verbal Spam?

2009-01-08T06:17:00.001-08:00

Toyota is rolling out a service called Lexus Insider that involves delivering audio messages of up to 3 minutes in length to your Lexus. Its an opt-in service available only on certain models. This could have some potential as a reminder service, but the potential for it to become annoying is very high.

What would be a useful service is one in which user specified audio be automatically aggregated in the cloud and then delivered to the vehicle whenever the user requests it. For example, if your favorite podcasts or music in your itunes library could be delivered to your car without having to flash to your media player and then hook your media player up to the cards audio system, that would actually be useful.

Read the article.

Tutorial - Writing a SOAP web service client in Java in 20 minutes or less

2008-12-29T11:46:00.000-08:00

So you don't know SOAP from WSDL? Whether you're coming from the RESTful school when it comes to building out web services, or you've simply managed to not get involved in web services altogether, here's a really simple tutorial describing how you can access an existing SOAP web service from Java in less than 20 minutes.

Preliminaries

To get started, we'll visit www.xmethods.net and find an interesting public web services that our Java client will attempt to access. For this tutorial, we'll settle on CDYNE's Free Weather Web service. Next we'll assume you have a recent vintage Java SE Developer Kit installed on your machine. Finally, we will also assume you have a recent version of Apache ant installed in your development environment, and that you are familiar building projects using ant.

Installing Apache Axis

You'll need to install a recent copy of Apache Axis. At the time of writing, we're using version 1.4. Axis is the Apache Web Services Project - an open source implementation of W3C SOAP and related specifications. It comes in both a C++ flavor, as well as Java so be sure to download and install the latter. Download the Apache Axis archive and unpack it on your local disk. We'll assume the root directory of your Axis installation is BASEDIR in the discussion that follows.

Generating Java Bindings for the Service

You'll notice that each public web service listed on www.xmethods.net has an associated WSDL description. Simply put, WSDL (pronounced "whiz-del") is an XML language that describes the interfaces of a web service. Since we want our Java client to access the CDYNE Free Weather service, we'll save a copy of its WSDL file to our local disk in a file named Weather.wsdl.

Apache Axis includes a convenient tool called WSDL2Java that lets you automatically generate the necessary Java bindings from a WSDL description. If you've included the jar files included in BASEDIR/lib in your classpath you can run this tool on your WSDL file as follows:

% java org.apache.axis.wsdl.WSDL2Java Weather.wsdl

We can also accomplish this with a simple ant build.xml file:

<project name="axisSamples" default="run-client">

<property name="librarydir" value="C:\axis-1_4\lib"/>

<path id="libraries">
<fileset dir="${librarydir}">
   <include name="*.jar"/>
</fileset>
</path>

<target name="run-wsdl2java">
<java fork="true" dir="src" classname="org.apache.axis.wsdl.WSDL2Java" classpathref="libraries">
   <arg line="../Weather.wsdl"/>
</java>
</target>

Either way, after successfully running the WSDL2Java tool you'll find the generated Java bindings for the services described in the WSDL file in the same directory. The generated classes will be found within a package hierarchy that was generated using the target namespace used within the WDSL file. In our example, the package will be com.cdyne,ws.WeatherWS.

Writing the Java Client

Next, we will write a simple Java client that utilizes the classes we generated from the WSDL file in the previous step. We'll begin by browsing through the generated code. Take a look at WeatherLocator.java. This is the class we'll use to bootstrap our client to get an interface to the Java stub that implements the interface to the remote web service. Looking at the class implementation, you'll notice this method:

   public com.cdyne.ws.WeatherWS.WeatherSoap getWeatherSoap() throws javax.xml.rpc.ServiceException {
  java.net.URL endpoint;
  try {
    endpoint = new java.net.URL(WeatherSoap_address);
  } catch (java.net.MalformedURLException e) {
    throw new javax.xml.rpc.ServiceException(e);
  }
  return getWeatherSoap(endpoint);
}

This method returns an instance that implements the interface com.cdyne.ws.WeatherWS.WeatherSoap. If we take a look at the interface (in WeatherSoap.java) we can see the details of the operations the SOAP weather service makes available:

public interface WeatherSoap extends java.rmi.Remote {

/**
* Gets Information for each WeatherID
*/
public com.cdyne.ws.WeatherWS.WeatherDescription[] getWeatherInformation() throws java.rmi.RemoteException;

/**
* Allows you to get your City Forecast Over the Next 7 Days,
* which is updated hourly. U.S. Only
*/
public com.cdyne.ws.WeatherWS.ForecastReturn getCityForecastByZIP(java.lang.String ZIP) throws java.rmi.RemoteException;

/**
* Allows you to get your City's Weather, which is updated hourly.
* U.S. Only
*/
public com.cdyne.ws.WeatherWS.WeatherReturn getCityWeatherByZIP(java.lang.String ZIP) throws java.rmi.RemoteException;
}

That's essentially all we need to know to write our simple Java client (TestClient.java) that accesses the remote web service:

import com.cdyne.ws.WeatherWS.*;
import java.util.Calendar;

/**
* @author spkydog (http://spkydog.blogspot.com)
*
*/
public class TestClient {

public static void main(String[] args) {

  try {
    Weather wservice = new WeatherLocator();
    WeatherSoap wsrv = wservice.getWeatherSoap();
    ForecastReturn retVal = wsrv.getCityForecastByZIP("49426");

    if (retVal.isSuccess()) {

      System.out.println("The weather forecast for "
        + retVal.getCity() + ", "
        + retVal.getState() +  " is:\n ");

      Forecast[] forecasts = retVal.getForecastResult();
      for(Forecast f : forecasts) {

        Calendar date = f.getDate();
        System.out.println(date.get(Calendar.DAY_OF_MONTH)
           + "/" +  date.get(Calendar.YEAR)
           + ": " + f.getDesciption());
      }

    }

  } catch (Exception ex) {
    ex.printStackTrace();
  }
}
}

Notice that WSDL2Java also generates all the various types used as input and output to the methods supported by the web service. In this case, we have a variety of types, of which we use two in the simple client above (ForecastReturn.java and Forecast.java)

Building/Running the Java Client

Once again, assuming that the axis jars are in our classpath, and our client and generated code is all in a sub-directory named src, we can compile our test client by entering the following command at the console:

% javac src/*.java src/com/cdyne/ws/WeatherWS/*.java

We can also build the code by adding a compile task to our build.xml file:

<target name="compile" depends="run-wsdl2java">
<javac srcdir="src">
<classpath>
<path refid="libraries"/>
</classpath>
</javac>
</target>

We can run the client by entering the following command at a commandline console:

% java TestClient

Or, if we haven't bothered to add the axis jar files to our environment's CLASSPATH variable, we can simply invoke the client via ant by adding this task:

<target name="run-client" depends="compile">
<java fork="true" dir="src" classname="TestClient" classpathref="libraries">
<classpath path="src"/>
</java>
</target>

If all worked as planned, the client prints out the weather forecast ranging from yesterday to six days into the future:

The weather forecast for Hudsonville, MI is:

28/2008: Partly Cloudy
29/2008: Mostly Cloudy
30/2008: Mostly Cloudy
31/2007: Partly Cloudy
1/2008: Mostly Cloudy
2/2008: Mostly Cloudy
3/2008: Mostly Cloudy

That's all there is to it! A zip archive with the complete source code and ant build.xml is available for download here.

TellMe's mobile directory service app wins WSJ Innovation Award

2007-09-24T06:35:00.000-07:00

TellMe's multimodal mobile directory search application won the Wall Street Journal's 2007 Innovation Award in the "Network/Internet Technologies" category.

Read the article.

Microsoft to Acquire Tellme

2007-03-13T07:35:00.000-07:00

According to folks in the know, Microsoft is poised to gobble up Tellme.

Read more from this ZDnet posting.

Some of the more interesting rumors are about how much its going to cost them.

Death by self-inflicted ASR wounds?

2007-02-01T17:37:00.000-08:00

George Ou over at ZDNet has posted an interesting article on his blog about a remote execution flaw when using Vista Speech Recognition. A posting on Slashdot referring to George's blog has generated and astounding amount of chatter on the issue, some of it rather humorous.

Turning off your PC's speakers unfortunately will not completely protect users from the vulnerability. Imagine a thug overpowering the conductor on your commuter train and simultaneously taking control of all notebook computers on the train running Windows Vista by giving commands over the train's PA system. Or, imagine millions of innocent teenagers being tricked into downloading rogue "trojan horse" ringtones that wreak havoc on all nearby rabbit-eared Vista machines when they received incoming phone calls! Or, what if some blind guy is using TTS (nope, can't turn those speakers off) to read this blog entry on his shiny new vista machine and he reacts to this sentence

"open explorer, delete star dot star, empty trash"

a couple of seconds too late? This vulnerability is a time bomb waiting to happen. ;-)

On the bright side, for us speech geeks, the fact that people are actually worried about this being a problem seems to suggest that speech recognition is getting pretty good. With modern up-to-date speech recognition software coming soon to +90% of the desktop computers near you, we're going to have to start handling our computers like our children - watch what you say when their around!

In the meantime, keep an eye on Richard's blog, he'll likely have something intelligent to say on this topic in the near future.

Speech recognition of unstructured human speech approaching 99% accuracy?

2006-10-04T06:52:00.000-07:00

The September 2006 issue of IEEE's Spectrum magazine has a rather interesting article reporting the results of a recent survey conducted among some 700 IEEE Fellows. The objective of the survey was to figure out what IEEE Fellows (not your average Saturday afternoon computer hobbyists!) expect or don't expect in science and technology over the 10 to 50 years.

When asked the question "Will computer speech recognition of unstructured human speech approach 99% accuracy?", 19.1% responded it was unlikely, 61.8% responded it was a likely. On the followup question "When is this likely to occur?" 25.2% indicated in 10 years or less, while 49.5% indicated 11 to 20 years.

The question itself is rather open ended and requires the respondent to make some assumptions. For example, are we to assume the unstructured human speech is coming from one or many speakers? Is the speech intended for human consumption, or is it assumed the user is speaking into a mic expecting a computer to transcribe to text? These sorts of issues make a world of difference in terms of raising/lowering the complexity of the problem.

In any case, its interesting to note that almost 20% of a crowd that one would assume consists of some of the brightest researches/engineers the world has to offer, responded negative to this question. Nevertheless, there's no need to fret. One can think of all kinds of examples where smart people were quite mistaken.

Perhaps the key take away here is the observation some of the more prudent fellows made: "science and technology are unpredictable."

Hands/Eyes Free on Windows Mobile 5.0

2006-10-02T09:43:00.000-07:00

Fonix recently announced VoiceCenter 3.1 for Windows Mobile 5.0 - coming to your smartphone soon for only 40 bucks! According to Fonix VP Walt Nawrocki, "speech recognition is a 'must have' to avoid tedious menus and button pressing." So just what is he saying here? Navigating the Windows Mobile 5.0 UI shell involves a bunch of tedious menus and buttons and badly needs a fix? Or, is there a deeper issue here... small mobile devices are inherently difficult for humans to interface with due to their small craniums and large paws?

spkydog would suggest in the case of Windows Mobile 5.0 there is a little of both going on here. One could argue that iPod user's aren't clamoring for speech interface, but than again iPods have fairly narrow functionality, while smart phones support a wide range of functionality.

Perhaps a more interesting question is whether or not the iPhone will need a speech interface... anybody care to make a guess?

Read about Fonix's VoiceCenter here.

Google's Mike Cohen on speech technology

2006-09-27T12:56:00.000-07:00

A recent podcast from ACM's QueueCast includes an interview with Google's Mike Cohen on speech technology. While this is more about Cohen's perspective on speech technology and less on where Google might be going with this stuff, its always interesting to hear what folks at Google are saying about speech technology.

You'll find a link to the mp3 here.

Microsoft discontinues...folds... integrates... unveils, and is right on...

2006-08-30T05:00:00.000-07:00

For those of you expressing concern that spkydog may have met an untimely demise while chasing the mailman, do relax. We've just been enjoying the waning dog days of summer while we can. Meanwhile, the speech industry keeps chugging along.

Earlier this month, tech journalists reporting on SpeechTek threw us what appeared to be a rather tasty bone when trying to interpret Microsoft's announcements at the conference. Initially, we just about fell out of our koop when reading the first line of an earlier version of this Information Week article which suggested that "Microsoft plans to discontinue selling speech server". After reading on, we understand that Microsoft is folding the Speech Server product into its Office Communication Server 2007 product line. Whew! We certainly wouldn't want to see Speech Server put in the pasture so soon after Microsoft put its weight behind VoiceXML.

There is a bit less hyperbole in the opening lines of this IT Week article, where the development is described as an unveiling of new speech technologies in Office Communication Server 2007. Nevertheless, the author dedicates at least half of the article to commentary on the now infamous speech demo Microsoft gave to financial analysts earlier this summer. Talk about beating a dead horse...

In any case, spkydog thinks Microsoft has made a wise move here, which will likely serve both them and the speech industry quite well. Despite the recent negativity and skepticism Microsoft's demo debacle has drawn, spkydog believes speech technology is more than ready for prime time - on our desktops, in our cars, on our PDA's/mobiles and of course our POTS phones, as usual. The bigger issue is getting the masses accustomed to using speech on a regular basis. Those of you who were introduced to using a mouse on your PC as an adult probably have memories of finding it rather cumbersome the first time. Speaking to machines has the analogous implications, and probably will for some time.

By integrating speech technology as a feature into systems that will be widely used by the masses (e.g., Windows Vista and Office Communication Server) Microsoft is putting speech technology in front of virtually everybody. Management at Acme, Inc. won't have to sit and spend cycles discussing whether or not its time to investing in a speech-enabled version of their favorite applications - its just going to happen. Selling speech technology as a stand-alone enabler is a tough sell.

There are actually very few companies who are in a position to market speech technologies to the masses in the way Microsoft is. Nuance has lots of speech technology, but in terms of apps is limited to the call center. Why buy Dragon when you get almost the same functionality bundled with Office? IBM has lots of apps and a fair amount of speech technology, but you run Microsoft bits on your PC all the time, and IBM bits only part of the time. Google certainly has the application reach, but their speech tech capabilities are not well understood, making them somewhat a wildcard.

In short, the folks in Redmond seem to be on to a decent strategy for differentiating their platforms and applications with speech technology.

Google to Release n-gram Models

2006-08-03T15:06:00.000-07:00

Today Google announced it will be providing a 6 DVD set of n-gram models it has generated from a training corpus of over one trillion words - culled of course from public websites.

"We believe that the entire research community can benefit from access to such massive amounts of data. It will advance the state of the art, it will focus research in the promising direction of large-scale, data-driven approaches, and it will allow all research groups, no matter how large or small their computing resources, to play together. That's why we decided to share this enormous dataset with everyone. We processed 1,011,582,453,213 words of running text and are publishing the counts for all 1,146,580,664 five-word sequences that appear at least 40 times. There are 13,653,070 unique words, after discarding words that appear less than 200 times."

Read the complete article.

More VoiceXML-Related Open Source Projects Announced

2006-08-02T06:27:00.000-07:00

Yesterday Voxeo announced www.rocketsource.org - an open source project involving three off-the-shelf VoiceXML/CCXML applications. It seems at least one of the applications (Voice Conference Manager) already existed as an open source project, but getting Voxeo behind certainly won't hurt. I would suggest the project is appropriately named, considering it's the rocket scientists over at Voxeo who are making it happen. I don't think the sun ever sets in that shop!

Read the press release.

Microsoft's Unfortunate SR Demo Takes First Place on Google Video

2006-08-01T06:04:00.000-07:00

For the past three days a 1 minute 39 second video clip of Microsoft's infamous speech recognition demo has the honors of being #1 on the Google Video Top 100 List. This is rather remarkable given the cacophony of often bizarre clips that it must compete with to earn the distinction. Please help me... is an unfortunate speech demo (alas they happen everyday... though not necessarily in financial analysts meetings) really as popular as Diet Coke & Mentos explosions, and lip syncing Chinese adolescents? Does anybody have any insight into the heuristics Google applies to generate the top 100 video list? One wonders if Google is simply having a little fun of their own at Microsoft's expense?

Watch the video.

Speech Recognition in Nintendo's Brain Age Not So Smart...

2006-07-31T07:38:00.000-07:00

spkydog came across this rather fun review of Nintendo's "Brain Age" game. The author's comments on this hit game are positive for the most part, until he gets onto the speech recognition functionality. spkydog has yet to play the game, but our current theory is that Nintendo might have licensed Vista's speech technology from Microsoft. ;-)

Just kidding on that last point of course! Microsoft's very own Richard Sprague makes a valid point in his posting regarding their recent demo debacle. To paraphase, people typically don't write about the 99 times the SR demo worked flawlessly. It is the failures that seem to be of human interest.

At the same time, unlike Vista, "Brain Age" is a shipping product on the DS. If the SR is really flakey on that platform (not surprising) it would of been better for everybody if they hadn't used the technology in this title to begin with. All it takes is a few high profile products that include less than acceptable speech technology to set up the negative stereotypes for yet another decade. If Microsoft does eventually get the speech technology right in Vista, it will be the most visible deployment of speech technology on the planet, and hopefully will do much in terms of helping the masses feel a bit less "buh-loo" about the readiness of speech technology in general.

Read the Brain Age review.

The Talking iPod?

2006-07-27T15:52:00.000-07:00

Earlier this month an article in the the Scotsman conjectures that a future version of the iPod will be equipped with a text-to-speech engine. They suggest TTS will be used to inform the user of what's playing when situationally impaired (driving, jogging, low light conditions, etc.). Evidently a recent patent filing with the USPO by Apple back in May has triggered this rumor.

Read the article.

Whoops... Microsoft's Speech Demo Bombs

2006-07-27T15:32:00.000-07:00

Evidently Microsoft was trying to convince some financial analysts that speech recognition was ready for prime time, and managed to prove just the opposite! No big deal really, this is classic for anybody who has tried to do a speech demo (or any demo for that matter!) in front of a crowd.

Read more on CNET.

Multimodal BoF at IETF Meeting this week

2006-07-11T08:40:00.000-07:00

A birds-of-a-feather (BoF) session is on the agenda of the 66th IETF Meeting in Montreal this week to discuss a draft charter of a working group to focus on the Distributed Multimodal Synchronization Protocol (DMSP). The protocol aims to coordinate multiple modalities in distributed multimodal systems and is described in an Internet draft submitted to the IETF by Motorola, IBM, and others.

The BoF is scheduled for Thursday, July 13 @ 1PM Eastern. If you're at the IETF meetings this week and interested in speech/multimodal you'll want to attend.

Read the DMSP BoF Agenda

Cool multimodal video in Google's top 100 List

2006-06-29T06:27:00.000-07:00

Here's a fantastic multimodal demo on a table top display in which the user uses a combination of speech and two hand interactions to control Google Earth and Warcraft III. I've seen Bill Gates showing some tabletop interaction scenarios at a keynote (CES??) sometime in the recent past, but this blows it away in terms of being a compelling demo.

The demo is based on the research of Edward Tse, a CS Ph.D. student at the University of Calgary. There is similar work being done at MIT.

You can watch the demo at Google Video.

Yet Another Voice Search Patent

2006-06-14T05:58:00.000-07:00

Yesterday V-ENABLE issued a press release announcing a mobile voice search patent granted by the USPO. spkydog hasn't yet had time to peruse the content of the patent, but it isn't clear that this is the first such patent in the mobile space as the press release claims. Google's recent voice search patent teaches a variety of client devices, including PDAs and telephones.

V-ENABLE's press release.

IBM Hires 100+ Speech Researchers?

2006-06-13T07:23:00.000-07:00

In the recent past we've posted and speculated about the fact that Yahoo and Google have hired a number of speech folks over the past year or so. This is the first I've heard about IBM beefing up big time in the speech area, at least in the recent past? Not sure who they might be, though we are aware of some who have left IBM recently. Hmmm...

The success of some of these limited-application voice recognition systems has recently prompted the big software heavyweights, Microsoft and IBM, to make further investments. IBM has hired more than a hundred extra speech technology researchers, with the aim of developing a system capable of matching the human level of speech recognition by 2010. And Bill Gates recently said that "we [Microsoft] aim to have computer systems capable of matching a human level of speech recognition by 2011"

Read the ZDNet article this quote was taken from.

Speech/VoiceXML Merger Mania - Where's Google?

2006-06-12T06:56:00.000-07:00

The Cisco announcement(s) last week add to the growing list of mergers/aquisitions involving vendors who are involved in the speech/VoiceXML industry. Here's the ones I can recall at the moment:

Cisco acquires Metreos and Audium
Genesys (Alcatel) acquires VoiceGenie Technologies
Cantata merges with Excel and Brooktrout which had acquired Snowshore earlier
Voxeo acquires Vocomo
HP acquires PipeBeach
Microsoft acquires Unveil and more recently picked up Vocalocity's VoiceXML technology
Genesys acquires Telera
Scansoft acquires Nuance, and many other firms.

I'm sure I'm missing some here, but absent from the list are Google and Tellme. Perhaps its time to resurrect the rumor of Google acquiring Tellme? :-)

Burger King Uses Speech Recognition to Take Drive-Thru Orders

2006-06-07T11:12:00.000-07:00

Here's a fun one for ya all... after firing their drive thru attendant and attempting to outsource the job of taking burger orders to a call center in India, this Burger King finally decides to give speech recognition a try.

Watch the video.

Prof. Hawking Reads Business Week?

2006-06-05T06:07:00.000-07:00

Here is a rather nice example of how the increasing degree of interactivity on the web makes rather boring redundant content a bit more interesting. In this recent Business Week technology column on speech recognition, the author regurgitates the same technology summary/predictions journalists have been writing for at least a decade now. The reader comments go on to add more interesting content/analysis than the original article itself. For example, if you think the guy in the cube next door who is constantly on a conference call annoys you, just wait until you hear the dull roar in the cubicles that is enabled when speech recognition gets as good as folks are predicting it will! Hopefully by then all of us technologists will be telecommuting!

One of the more interesting comments is supposedly supplied by Prof. Hawking's himself. Not sure if that's the case, but the point made (conventional ASR does a pretty good job of recognizing conventional TTS) is actually something I've confirmed independently. Unlike Hawking's, the typical use case for most of us is to automate testing of speech systems/applications.

Read the article.

MIT's Human Speechome Project

2006-05-30T05:30:00.000-07:00

For those of you speech folks out there with kids, you've likely never ceased to be amazed and inspired by the fact that your 3 year old child's speech recognition capability surpasses the most advanced speech technology available. Researchers at MIT are trying to figure out just how these little tykes get so good with speech. The team will be monitoring all of a child's entire first 3 years with sophisticated video cameras, etc. in an attempt to understand how he acquires its speech skills.

I wonder if the child will have access to the videos later? Imagine watching yourself learn how to say "mama"? I've noticed with my own children that when they watch videos of themselves as infants, they tend not to be able to filter what they actually remember experiences vs. what they saw on video when they were older. For example, my daughter will say "remember the time I was wearing Daddy's sunglasses in my playpen?", when in fact she doesn't remember it at all, she just remembers seeing a video of it later.

In any case, like a lot of the projects at MIT, this is a fascinating study, and it will be interesting to see what it turns up in the years ahead.

Read the article.