自然语言处理、语音信号处理等信息技术的发展使得对话交互界面（Conversational User Interface，CUI）获得广泛使用，语音助手、智能客服等都融入了CUI的设计。然而，目前的CUI设计还有很大的改善空间，例如：1）受限于市面上已有产品的语音交互形式，缺乏创新性和多元性；2）更多从技术角度出发关注易用性，而忽略用户体验中的其他因素，例如对话中的情感；3）应用实践和理论研究有较大的脱节，相互之间的联系远远不够。因此，本文旨在探索如何为具有创新性的多元化人机对话交互提供一套切实可行的设计方法。本文首先对CUI相关的学科领域进行了文献调研。随后对CUI的特点进行了分析，并归纳出自然人机对话的三方面特性，即系统性、多元性、不确定性。基于情感化设计理论，重点提出在CUI设计中进行层次化用户体验的概念，即在设计时考虑直觉感受层、对话功能层、对话认知层三个层次的用户体验。在此基础上，本文提出与层次化用户体验概念相结合的CUI设计方法，包括CUI交互流程框架和设计流程，流程强调在多个环节考虑多层次用户体验。本文强调选择可以促成人机对话形式创新的设计变量进行研究，并进行了以下两项具体场景中的CUI设计研究。首先，聚焦智能访谈的场景中CUI的追问行为设计，深入分析了追问行为设计对多层次用户体验的影响。然后，在CUI作为任务助手的场景中，对其情绪感知和反馈行为的设计进行研究，并深入分析情绪反馈行为对多层次用户体验的影响。两项研究通过基于对话原型的用户实验发现，设计方案会在不同的用户体验层次产生不同的影响，因此提出了相应的优化设计方案。上述两项研究均以本文所提出的CUI设计方法作为指导，为设计师在CUI设计时考虑多层次用户体验提供了一定的参考价值。其中，主动向用户追问与感知用户情绪并反馈的设计探索，在人机对话形式方面具有一定的创新性，不但对本文所提出的CUI设计方法进行了具体场景下的实践检验，还为多元化人机对话交互设计提供了理论研究和实践应用的参考。
Information technologies, including natural language processing, speech processing, etc., enable conversational user interface (CUI) in various application scenarios. For example, voice assistants, customer service chatbots, and other popular applications commonly use the CUI. However, current CUI design: 1) is constrained by the voice interaction of existing products and lacks diversity and novelty; 2) focuses on usability instead of holistic user experience; 3) needs more systematic research both in design theory and practice. Thus, this thesis aims at providing a design method for the novel CUI with diverse conversational interactions.Based on the literature review, this thesis presented three properties of the natural human-computer conversation: uncertainty, diversity, and systematicness. As the primary theoretical contribution, the thesis proposed the concept of holistic user experience in the conversational interaction based on the Emotional Design Theory, which can be briefly described as considering the user experience from three perspectives: intuitive feeling, conversational behavior, and conversational reflection. Further, the concept was integrated with the interaction framework and the design process of CUI. Specifically, design variables in CUI were analyzed in the three aspects of user experience, and the user study for CUI design highlighted its evaluation of user experience from multiple perspectives.Meanwhile, this thesis explored novel forms of human-computer conversation with two studies. The first study focused on the CUI's follow-up question design in the interviewing scenario and investigated its effects on the user experience. The results of the user experiment showed that the CUI's follow-up question design supported the user's information sharing and ensured the relevance and fluency of the interview. The second study concentrated on the CUI design of emotion perceiving and responding in the scenario of task assisting and tested its effects on the user experience. The results of the user experiment showed that the CUI design of emotion perceiving and responding enhanced the user's Perceived Emotional Intelligence ratings on the CUI and alleviated the user's negative emotions caused by the tasks. Further, the results of both studies revealed that different designs, i.e., different types of follow-up questions in the first and various types of emotional responses in the second, affected user experience from different perspectives, which provided insights for the future CUI design. The presented studies provided empirical evidence that the proposed CUI design method, including the interaction framework and the design process, was practical in the related scenarios. The computer actively asking follow-up questions and responding to user emotions were novel forms of conversational interaction. Therefore, the presented studies shed light on the diverse human-computer conversational interaction.
