Talking to computers - voice recognition as a mode of interaction for executives
Amiram Mel · University Microfilms International eBooks · 1983
Voice recognition systems are used today to perform tasks in situations where there is a need to free the hands and/or eyes, or in a very special environment such as total darkness, weightlessness, etc. Recent developments of new technology have prompted us to determine whether voice can be used not only for such unusual situations but also for tasks currently accomplished by other input media such as the keyboard. Our research was conducted in the office environment, a common place where people work and where voice is the major communication channel. Once the feasibility of using voice as an interaction mode with computers was demonstrated the issue of preferability explored. We discovered when and why users prefer to use voice rather than keyboard. We conducted various types of experiments to cover all aspects of the voice recognition process. We demonstrated that voice was preferable over keyboard for spreadsheet packages (i.e. VISICALC, SUPERCALC) especially for data manipulation (i.e. sensitivity analysis, what if analysis). Users preferred voice for these types of tasks because they could watch the screen continuously. Keyboard usage required head movements between the screen and the keyboard. Special effort was devoted to the training stage. Several training modes were tested and we found that the random mode of vocabulary organization and vocabulary presentation is the best mode. We concluded that an adaptive training method, which would update the representative voice pattern, not only during the training stage but also in the recognition stage, would improve the accuracy and the performance of this current technology. Our experiments also illustrated indirect advantages that using voice can give the users. Voice interface allows users to operate similar packages with the same set of words therefore eliminating the need for additional and special training. It also allows users to interact with computers in several natural languages.