Dialogue Processing for Voice Interactive Systems

这项研究是与杜克大学的艾伦·W·比尔曼合作进行的，涉及如何通过语音与自然语言的计算机系统交互的问题。它继续并建立在调查人员在过去七年中进行的研究结果的基础上。在电子设备维修的问题领域工作，他们目前正在研究一系列的交互模式，从“被动”(计算机只是听从命令并回答问题)到“直接-指令”(它负责、提问并指导人类用户)。他们还在研究任务导向型对话中成分的性质和作用，以及如何表达和生成适当的、完全语法的、自然语言的SPO-KEN回答。这项研究的激励目标是能够设计一个机器伙伴，能够进行自然语言语音对话，并能够在电子维修任务中帮助不同能力和经验的人类。在实现这一目标方面取得的进展，在很大程度上造就了如何使用自动机的知识，而这些自动机以前只有熟练的人类才能发挥作用。

This research, undertaken in collaboration with Alan W. Biermann of Duke University, is concerned with the problem of how to int- eract with a computer system in natural language by voice. It continues and builds on the results of research carried out by the investigators over the past seven years. Working in the problem domain of electronic equipment repair, they are currently investigating a range of interactive modes from "passive" (the computer merely follows orders and answers questions) to "direct- ive" (it takes charge and questions and instructs the human user). They are also investigating the nature and role of com- ments in a task-oriented dialog and ways of representing and generating appropriate, fully grammatical, natural language spo- ken responses. The motivating goal of the research is to be able to design a machine partner, capable of natural language voice dialogue, and able to assist human beings of varying abilities and experience in an electronic repair task. Progress toward this goal contrib- utes in important ways to knowledge of how to employ automata in roles previously possible only for skilled human beings.