近日，实验室博士生韩钦宇关于对话文本摘要生成的研究成果被期刊IEEE Transactions on Audio, Speech, and Language Processing (TASLP) 录用。TASLP期刊是音频、声学、语言信号处理的顶级期刊，致力于研究处理音频、语音和语言信号的创新理论和方法及其应用，为中科院一区SCI期刊，在CCF学术推荐列表中认定为B类刊物，清华最新版计算机学术推荐列表中认定为A类刊物。

题目: Let Topic Flow: A Unified Topic-guided Segment-wise Dialogue Summarization Framework

摘要: Dialogue summarization is a task that aims to condense dialogues while retaining the salient information. However, due to the different domains involved in the dialogue, the corresponding format of reference summary varies from each other, e.g., QA pairs for customer service and SOAP notes in the medical field. To address the common challenges encountered in various fields and alleviate differences in the generation process, we introduce a novel unified topic-guided dialogue summarization framework, by which we can first capture the topic structure of one conversation and use it to guide the generation of summaries. This framework is the first to model the fine-grained topic structure of the dialogue and pose its identification as a Seq2Seq task, as well as introduce the topic-guided segment-wise attention to produce the final summary in segments following the specific format in each domain. Such a concise but effective method avoids the trouble of customizing decoding schemes while retaining the topic structure of dialogue in its summary as much as possible. Comprehensive experiments were conducted on four benchmark datasets in different domains and the results show the better performance and generalization of our method compared with the baselines.

对话摘要生成旨在从对话中提取关键信息并整合生成简洁易懂的摘要。由于对话涉及到的领域不同，其相对应的摘要格式不尽相同，如在医学领域，摘要通常以SOAP的格式体现，而在客服场景下，摘要一般以问答对的形式体现。为了处理不同场景下摘要生成过程中遇到的普遍的问题并且统一不同格式摘要的生成范式。我们提出一个统一的主题引导摘要生成框架。在这个框架下，我们首先提取整个对话的主题脉络并以此作为指引生成最终的摘要。值得一提的是，本工作是第一个提出构建细粒度的主题架构并且使用端到端的方式生成主题框架。同时引入分段式主题引导注意机制来来辅助主体框架生成最终摘要。这样的一个简洁高效的框架可以从对话中精准的总结出主题脉络并避免为不同格式的摘要制定不同的生成范式。在四个数据集上进行的大量实验证明了我们所提出的模型的有效性以及在不同领域任务上的普适性。