The role of audio effects in interactive voice response systems

摘要

Interactive voice response systems, which use speech synthesis and speech recognition over the telephone, are becoming popular. Such systems are able to connect to the Internet, so any telephone will be enabled to access Web-based services via voice browsers. Using the current technology, however, it is difficult to recognize and understand the user's spontaneous speech, so it is important to guide the appropriate usage of speech input to the user. In this work, we investigate the difficulties in using voice response systems, based on the Norman's model of human-interface activities. We then propose some usage of audio effects in the outputs of dialog systems, which assist in overcoming the difficulties. In addition, we have built an experimental voice response system and have evaluated our investigations with novice users of the system. Our proposed method results were shown to be effective in reducing the mistaken use of the vocabulary and inappropriate barge-ins.