As a product it's a great idea, I'm sure it would sell very well. As an effective learning tool I'm not so sure. Unless you could implement top notch speech rec

As a product it's a great idea, I'm sure it would sell very well. As an effective learning tool I'm not so sure. Unless you could implement top notch speech recognition it would not do too much in improving your speech. And so far I'm yet to see English speech recognition software that would be suitable for such a purpose. The situation is even worse in case of Mandarin (and probably Cantonese too). Mandarin speech recognition that is out there is not suitable for anything more than simple voice-commands, at least last time I searched. Generally speaking I'm not an advocate of practising speech before attaining a certain, relatively high level of proficiency. But that's a completely different matter. A text bot from a technical standpoint would be completely doable and should be a good product as well. You could actually build a pretty sophisticated course utilizing that text bot, a lot of people would be interested I'm sure.