Spoken language technology research at Microsoft

Abstract

The goal of ongoing work is to develop spoken language technologies for use by a broad audience of computer users and in a wide range of applications. The current work in support of this broad objective falls into four main areas: speech recognition technology code-named Whisper (Windows Highly Intelligent Speech Recognizer); speech synthesis technology code-named Whistler (Windows Highly Intelligent Stochastic Talker); spoken language understanding technology code-named Leap (Language enabled applications); and the Microsoft Speech Software Developer’s Kit that facilitates full integration of speech into applications. This paper describes ongoing R&D effort and technical challenges in these areas. Not only are technology components being created that can be shared for the whole spoken language system, but also solutions are being created for application developers to author compelling speech-enabled applications. To realize this vision of speech enabling the whole computer, not only should critical technology components be improved (i.e., Whisper, Whistler and Leap), but also a tighter integration of these spoken language technology components should be evangelized with broader applications and interfaces to enhance and extend the entire user experience. A number of groups have started developing SAPI-compliant technology components and applications such as telephony, wireless PDA, accessibility, and desktop agent. More information can be found in http://research.microsoft.com/stg.