Workshop on Speech Summarization

语音文档(例如，日常会议、广播新闻、讲座、政治演讲)的自动摘要为这些数据源中的信息访问提供了一种有效的方式。针对不同领域的语音数据，如会议、广播新闻和演讲，已经开发和评估了不同的摘要方法。然而，与文本摘要相比，语音摘要并不成熟。没有基准测试，评估指标不明确，可用的带注释的数据集也不多。由于最近社会对语音摘要的兴趣与日俱增，有必要召集该领域的研究人员来讨论这项任务中的各种问题和未来的研究方向。这次研讨会聚集了致力于语音总结各个方面的研究人员，帮助为这项任务建立一个研究社区，并促进关于开放问题、未来方向、数据收集和共享、算法和结果的讨论。讨论的主题包括语音摘要的定义、数据收集和注释、数据共享、评价指标、可能的共享任务、领域效应、摘要方法、文本和语音摘要的异同以及人类交际中言语以外的情态。本次会议的讨论将通过一个专门的网站进行传播。这次会议将极大地推动演讲总结研究的发展。参加这次会议的学生将从讨论中受益，并能够在未来的研究中使用这些讨论。

Automatic summarization of spoken documents (e.g., everyday meetings, broadcast news, lectures, political speeches) provides an efficient way for information access in these data sources. Different summarization approaches have been developed and evaluated for speech data in different domains, such as meetings, broadcast news, and lectures. However, compared to text summarization, speech summarization is not as mature. There are no benchmark tests, the evaluation metrics are not clear, and there are not many annotated data sets available. Because of the recent increasing interest in speech summarization in the community, there is a need to gather researchers in the field to discuss various issues in this task and future research directions. This workshop brings together researchers dedicated to speech summarization in various aspects, helps build a research community for this task, and fosters discussions about open problems, future directions, data collection and sharing, algorithms and results. The topics discussed include the definition of speech summarization, data collection and annotation, data sharing, evaluation metrics, possible shared tasks, domain effect, summarization approaches, similarity and differences between text and speech summarization, and modality beyond speech in human communication. The discussions from this meeting will be disseminated using a dedicated website. This meeting will greatly help move forward the research in speech summarization. Students participating in this meeting will benefit from the discussions and be able to use them in their future research.