Almost every auto manufacturer has voice recognition-based applications in their high end models, but the technology has always been suspect. Even Ford's Sync technology, while popular, does not use natural language techniques and requires highly structured command & control utterances from users.
Contrary to conventional wisdom, I believe that natural language voice recognition is finally getting to a performance (memory usage, compute cycles usage) and accuracy level that makes it feasible to use inside cars. Lots more innovation is coming down the pike - in addition to the incumbents Nuance and IBM, there are many other vendors in the voice space - from the brute-force/off-board approach of TellMe to the hybrid off-board/on-board approach of Vlingo to the natural language approach of companies like Speak with Me.
MIT's Computer Science and Artificial Intelligence Laboratory is building an interesting natural language voice recognition system. Their City Browser is a "research prototype which lets you speak and draw naturally to obtain information about restaurants, museums, subway stations, and addresses."
Below is an image of what came up when I said "Show me a map of Boston."
Comments