Last week I read about a product called vlingo in the New York Times. It strikes me as an important technology development. 

vlingo is voice recognition for smart phones, the Blackberry to start. With it, I can speak a command (e.g., “open calendar”), dial by voice a phone number in my contact list, or dictate an e-mail or SMS message.

Two aspects of vlingo technology are interesting. First, all voice recognition processing is server side: vlingo transmits your speech (via standard audio files) to its servers, where its advanced software performs voice recognition. In essence, this makes the smart phone a “thin client” because the heavy duty processing is performed remotely.

If you think about this in connection with “cloud computing”, we may not be that far “an office on your smartphone.” (Voice recognition and Google apps might be all many people need for basic personal productivity needs.)

vlingo solves one of two big problems with hand-help devices: input. As a fast typist, I’ve never been a dictation fan but I cannot type nearly as fast with just two thumbs as I can with 10 fingers doing touch typing. So I plan to keep using vlingo.

The other big problem is screen size. We need folding or rolling screens. With organic light emitting diode screens now in commercial production, that day may not be so far off. With both, the computing world could see a dramatic shift.

I might at one time have said local computing power is a big problem. Between advances in chip and the vlingo demonstration of client-server computing, I don’t think that is a constraint. (Of course, mobile carriers must provide sufficient bandwidth.)

The second intriguing aspect of vlingo is the approach to voice recognition. vlingo’s uses Hierarchical Language Models and adapts to users and applications. I wonder if these techniques could be used to help speed reviewing documents in e-discovery.