Teaching language fluency

I like the idea of using SR technology for purposes beyond simple recognition or synthesis, and I've already mentioned one of my favorites: foreign language instruction.  Now I just heard about Teaching Machine, and their "SpeakIt Suite", which they describe as "a cost efficient solution to perfect oral English".  Here's what they're presenting at The Venture Forum in San Francisco on May 1st and 2nd.

Teaching Machine, Inc. - Software
Teaching Machine is a California corporation with elegant, interactive, voice-activated, multimedia software to teach language fluency. The first product, for teaching English reading and speaking uses a sophisticated speech processing technology that can “listen, analyze and correct” like a human teacher. An artificial intelligence approach combines a) the customization of the software for use by adults and children with a wide range of proficiency, b) an automated content upload system that converts any published content into curriculum, and c) a real-time comprehensive performance assessment system. The core innovation lies in the product's scalability, allowing for “one product, many markets” concept.

I don't know anything about this company, so it's hard to say how it works.  It would be pretty easy to build something with Windows Speech Recognition that can recognize phonemes and tell you whether you pronounced them correctly or not. The hard part, as always, would be in the usability of the software -- making it seem natural, rather than a simple rote-and-repeat exercise that would quickly turn boring.