Predictive Text + Speech = Improved Input for Mobile Search

The $265 million that Nuance Communications is paying for Tegic is a sign that predictive texting (like Tegic's T9 protocol) can complement, rather than compete with, automated speech-based dictation. When the leading automated speech processing technology provider makes a point of acquiring a company who's core product had long been trying to make dictation systems unnecessary, there's obviously more to it than meets the eye.

Today, a three-year-old company called TravellingWave came out of stealth mode to announce that it is fast approaching the point where it will formally introduce embedded speech processing software that supports "predictive speech-to-text" translation. Ashwin Rao, the founder of TravellingWave, told us that, while other embedded speech recognition software vendors position their products as "keyboard replacements", TravellingWave sees it as a keyboard enhancement. Thus the stage is now set for solutions providers to support more efficient and speedy entry of data or text in the form that is most appropriate to the task at hand.

TravellingWave sees dictation of text messages and e-mail, as well as the input of mobile search terms, as core application areas. The product is at a "very advanced beta stage" right now. Today's press release was to serve notice that it had received funding from a sophisticated set of angel investors with experience at McCaw Cellular, Microsoft, Nextel and Google.

Predictive speech-to-text is a subtle enhancement to multimodal communications. Classic xHTML+Voice Profiles (X+V) based of multimodal applications let Web-based services treat spoken words like text-input. Most instantiations use speech for input and render prompts or system responses as text, graphics or Web pages. Predictive speech-to-text allows for users to toggle back and forth between modes of input using keyboards or keypads if necessary or appropriate to overcome noisy conditions or protect privacy.

According to Rao, the technology is designed to "build on what people use already." The intellectual property is built around handling tens of thousands of of words in ways that support both predictive typing and spoken utterances. The beta software is a stand-alone speech-to-SMS application. According to company spokespeople, users "have the impression that it would have taken them a much longer time to finish their tasks if they hadn't used the application."