This press release from IBM claims that speech recognition for the Web will be a hot technology within five years. As I mentioned previously, IBM has been dedicating some effort to speech recognition. It's a nice idea, I'd like to see it happen, but it's usually a good idea to be skeptical of claims for any speech recognition application.
Here's a slicker presentation on the material in the press release, with a little different spin. The "web" use would be for automatic speech translation, apparently. Automatic translation has been a dream for decades, and in fact it's one of the original AI research projects. I'd like to get a chance to talk to someone in IBM speech research about this.