Voice Command over the Web

Question

I'm just interested in this scenario. Suppose you are building a web application, and would like to receive speech commands and execute actions based on the speech-to-text result.

Any major roadblock that may otherwise make this a killer application?

Answer 1

The first thing I can think of is probably the quality of audio transmitted will largely depend on upload bandwidth of the user. The less bandwidth, the less quality, the less able the speech-to-text algorithm can detect what is exactly being said.

Answer 2

Low sample audio can be used for s2t algorithms. My problems:

1. Getting the audio to the server and processing it (flash or Java or something)
2. Having the client poll for the required action
3. Making this scalable on the backend
4. Dealing with the wide range of languages, dialects, inflections and accents that the Internet supports
5. Ensuring it is unobstructive and that a usable fallback is available
6. dealing with complaints from usability people

There are tons more im sure, but other than that go for it.

Where would we be without people going "we are going to the moon" and then doing it. Go for it ...

If it hasn't already been done, if you fail or succeed then you will probably learn something cool.