Methods for formatting text produced with a speech-based editor

Abstract

The large number of typists and secretaries in the work force who type documents for others are evidence that a substantial segment of the population can not or will not produce their own documents. There are, however, significant pressures in the work place for shifting some or all of the responsibility for document production to the producer of the document content. Automatic speech recognition systems would make this possible if they were to provide a means for handling formatting operations as well as transcribing content accurately. This study investigated user reactions to six formatting methods (e.g., for creation of a list structure, for beginning a paragraph) in a simulated 'document creation by voice' application. The results of this study indicate that subjects preferred the method in which a spoken command was used to produce formatting and immediate recognition feedback was received. The second highest ranked condition was one which required that, in addition, the spoken command be preceded by a mouse click. An argument is made in favor of implementing the second-most favored method because it would allow for the inclusion a a number of additional features requested by users.