Defense committee: Dr. Matthew Stone, Dr. Alex Borgida, Dr. L. Thorne McCarty

Abstract:

Disambiguation has the potential to be a major problem for speech recognition systems. Since many English words have multiple meanings, there is a need to disambiguate the words to determine user intent and respond correctly. Through the use of rules and user feedback, the disambiguation problem can be resolved. Since the correct action would be executed, this method would result in higher user satisfaction.

The usage of the concepts of iterative design and designing for error are also essential for the development of quality, user-friendly speech recognition systems. These concepts were utilized in the development of the Personal Assistant program which was used as the case study in this research.

The case study implements a general disambiguation class. Through the use of a series of methods, the application is able to determine the appropriate questions to ask and actions to take in order to interact with the user to find the desired solution. This set of methods can be used by any application to determine the user's intent with minimal changes. By interacting with the user and utilizing the current state of the application, the disambiguation problem in the Personal Assistant program was successfully resolved.