Wednesday, September 05, 2007

Speech-to-text using your cellphone

Read/Write Web has posted a review of Jott, a mobile phone service that allows users speech-to-text functions such as dictating email messages. They have just released a service called “Jott Links” (and an API) that allows users to interact with websites like Zillow and Twitter using voice commands and speech recognition.

On the one hand this seems like a no-brainer, the perfect application of this technology. I spent a summer during college working for Speech Technology Magazine, and back then speech recognition software was mainly marketed for people with motor disabilities and those who liked to dictate their texts. With the proliferation of cell phones, it seems like this deployment of the technology should be a perfect fit for most people’s lifestyles. However, the limitations of the technology make me wonder how it will work. I could be hopelessly out of the loop here, but the speech recognition programs that I was familiar with back then depended on creating profiles of individual users, slowly learning how to decode the user’s speech through a trial-and-error process that required a lot of feedback in the form of corrections. Will Jott’s service do this, or has speech recognition evolved beyond this problem? If not, it could be a serious drawback. If a user wanted to post a message to a public service like Twitter, he or she will certainly want to make sure that message doesn’t contain any embarrassing malapropisms.

It will be interesting to see if this feature catches on. There have been some significant developments in interface design lately, most notably the huge response that Apple has received for the iPhone and the iPod Touch, which was just released today. Perhaps they will coexist, and we will get used to an everyware-esque situation where our interactions with computing devices will fit much more naturally in our everyday actions.

No comments: