Spoken Language Operating System

This SBIR program seeks to develop a spoken language based computer operating system that performs ordinary office activities including menu navigation, application start up, E-mail, word processing, spreadsheets, etc. The objective is to provide a spoken language alternative to the keyboard. The application places stringent demands on vocabulary size, speaker independence and domain robustness that exceed the capabilities of commercially available speech recognition technology. The approach proposed by this program is to use shallow Natural Language Processing (NLP) and extensive lexicons to supplement the traditional Hidden Markov Model approach to speech recognition to help achieve the required performance. This will be accomplished using a variety of very practical enhancements to existing speech and natural language modules that are available commercially or through the research community. The important role of the dictionary as a source of semantics will be exploited extensively. The proposing contractor has succeeded in applying a similar approach to enhancing text processing with NLP that resulted in a recognized commercial product.