The SpeechWorks division of ScanSoft has now been involved in over 1000 deployed speech applications worldwide. These systems have been deployed in a number of environments including telephony call centers, automotive systems, and mobile devices.

In this talk. I will describe the evolution of those systems over the past 8 years. During this time, we have gone from very constrained directed dialog systems, to systems using statistical language modeling, to systems supporting mixed initiative dialogs. Also during this time, there has been a corresponding progression of development environments and tools to support these systems. I will also cover the progression of standards, including VoiceXML and MRCP and some new efforts to standardize dialog descriptions and discuss the runtime environments needed to support more advanced dialog features.