Natural Language Prompts for an Automated Call Router: Examples Increase the Clarity of User Responses

Abstract

Speech-enabled interfaces are proving to be an effective option for service-center call routing applications. However, the effectiveness of a call routing application is dependent upon the speech recognizer correctly interpreting the caller's utterance. One approach to increase the clarity and routability of callers' initial requests is to provide examples within an open-ended prompt to provide context and guide caller's speaking behavior. Two experiments were conducted to investigate the effect of including examples in the opening prompt for a natural language call routing application. In both experiments, callers were asked to say the purpose of their call with an open-ended prompt. Half of the caller's were presented a prompt that contained examples, and half of the callers were presented with a prompt that did not contain examples. The results showed an advantage for the use of examples in terms of the percentage of routable caller utterances. The design implications for natural language prompts are discussed.