毕业生知识库

人机对话系统是指人类用户通过自然语言与计算机进行对话的交互系统。整 个系统通常由多个模块组成，其中对话理解和生成是人机对话系统中关键的组 成部分。近年来，随着人工智能技术的蓬勃发展，对话理解与生成技术的研究取 得了较为丰硕的成果。然而，已有的工作主要关注常规领域的对话理解和生成方 法，对医疗领域关注较少。医疗领域的对话理解任务面临着对话表达口语化、信 息冗余等诸多问题。此外，在医疗领域中，已有的情感陪护系统大多只适用于处 理简单的情感类别，缺乏对细粒度情感的建模。针对上述不足，本文围绕医疗领 域对话文本的建模、上下文选择和信息融合方法展开，旨在改善已有方法的上下 文利用不足的缺陷，使模型能够更好地对上下文进行建模和利用，实现对话系统 性能的提升。

Human-machine dialogue systems use natural language to communicate with humans. Typically, it consists of multiple modules, of which the  understanding and generation modules are key components. In recent years, there has been significant progress in the dialogue understanding and generation tasks. However, most existing work mainly focuses on conventional fields and pays less attention to the medical field. Dialogue understanding and generation in the medical field still face several challenges, including conversational characteristics like colloquialism, redundancy, and so on. Fur- thermore, most of the existing dialogue generation models of emotional chat systems are only suitable for dealing with simple emotion categories and lack fine-grained emotion modeling. To address the above problems, this dissertation focuses on the approaches of modeling, selecting, and incorporating information into dialogue systems, so as to alleviate the deficiencies of context utilization in existing methods and make better use of contextual information to improve the quality of the dialogue system.