Testing the effect of training with synthetic speech on task performance with a mixed-human-and-synthetic-speech interface

Applications that use human speech for fixed prompts and synthetic speech for dynamic content are commonly used in commercial speech applications because of the infeasibility of pre-recording dynamic content with human speech. Users generally have a more positive attitude toward speech interfaces that mix human and synthetic speech than those that use only synthetic speech. However, user attitude in using mixed speech interfaces is not necessarily correlated with user performance. Studies have found user task performance to be significantly better with synthetic-speech-only interfaces than with mixed-speech interfaces. I conducted an experiment to determine if training with synthetic speech would improve user task performance with mixed speech. I predicted that this would improve users' familiarity with synthetic speech and thus would improve their performance on tasks using mixed speech. I evaluated two groups' performance on a task using a mixed-speech interface: users given a prior task with a synthetic-speech interface and users given a prior task with a mixed-speech interface. The results of the experiment showed no significant difference in task performance between the two groups as measured by written answers to questions and number of repetitions of messages. In addition, there was no significant difference between the two groups on self-assessment of task performance or liking of the synthetic voice. The results indicate that users' lack of familiarity with synthetic speech is not a major cause of their lower task performance with a mixed-speech interface.