Around now Microsoft Agent technology is coming up to its first birthday. So far it hasn't created so much as a spark, not even a fizzle of interest. Which is a shame because it has the potential to transform Webpages into something much better, namely voice-activated user interfaces. Instead of having two-dimensional Webpages, we now have voice-enabled Web applications and the Internet takes another one of those characteristic quantum leaps.
If you have never come across them - and most people haven't - Microsoft's Agent technology has nothing to do with Web agents which trawl the Web looking for information (see Computimes of July 13th). Microsoft Agents are three characters: Merlin, Genie and Robbie the Robot. They are very similar to those animated agents in Office 97, but a lot less irritating. Taking the agents out of the Office suite and onto Webpages lowers the cringe factor; control is placed into the hands of a Web designer and out of the hands of the people who brought us "Bob" - a computer buddy for novice Windows 95 users that was the butt of many a joke.
Microsoft and others have been working on voice-enabling applications for a couple of years now in the desktop environment but so far there is very little on the Web. Over the past 12 to 18 months a number of excellent packages from Dragon and IBM (Via Voice), as well as from Microsoft, have, for the first time, allowed us to talk to MS Word and other applications in order to perform simple tasks like dictating a letter. Microsoft Agents is one of the first attempts to move this voice-interaction to the Web.
The technology behind Microsoft Agents is the excellent Lernout and Hauspie's TruVoice engine which Microsoft licenses and also plans to incorporate into future operating systems. There is no real voice training required. This makes it perfect for the Web with its myriad of different users and accents. Nothing more than a cheap microphone and a soundcard will allow you to get around Microsoft Agent-friendly Webpages. I would be lying if I said it was 100 per cent accurate at detecting the spoken word, but it is pretty close.
I recently added a simple voice-activated pop-up navigation menu to my own pages. You can try this out at www.riis.com/msagent. Each animation has a number of possible movements which can be easily coded or scripted in VBScript, JavaScript, C++ or Java. Other freely available tools allow you to create custom characters with your own or different voices. The standard voice engine is very reminiscent of Stephen Hawking's voicebox.
Microsoft Agents can be used to allow hands-off navigation around a Website - especially useful for someone with arthritis or RSI who finds it difficult to use a mouse or keyboard. It would also be possible for blind people to talk to a Website and for the Website to reply. However, because there are so few Microsoft Agent Websites around, it is highly unlikely that any really useful sites or gateways exist as yet. This is a real pity as the current crop of Braille-enabled browsers are very limited. (see Computimes of 27th July for an account of browsers for the visually-impaired).
So why aren't there more Microsoft Agent Webpages? Well the downside, and there always is a downside, is that the three core engines - the Lernout & Hauspie software which converts text to speech, the Microsoft Command and Control engine which analyses your speech and the Agent software itself require a 5MB download. The other significant disadvantage is that MSAgents are ActiveX controls; don't expect to see anything if you don't use Internet Explorer. I had hoped that future versions of Microsoft Agents would become an intrinsic part of Internet Explorer which would solve the download problem. But this is very unlikely given the distinct lack of interest in the Web developer community, even for use on local Intranets where the type of browser can be controlled.
Because there are so many programs and plug-ins on the Web, occasionally something very worthwhile is overlooked. You can check out the Agents at www.microsoft.com/workshop/imedia/agent and, if you have a fast link, do try my demo.
Godfrey Nolan is at godfrey@riis.com. See also www.riis.com