A visit to Stanford University, California, USA

经过一系列的会议，斯坦福大学的Stanley Peters教授邀请Whittaker和他的研究小组于2008年4月/ 6月在斯坦福与Peters一起工作。他们将共同研究如何从会议记录中智能提取信息。计划中的工作扩展了两个小组已经开展的相关但不重叠的研究（Kalnikaite and Whittaker, 2007, Purver et al., 2006, 2007, Tucker and Whittaker, 2006, Whittaker et al., 2007, in press）。目标：-目前在英国，欧盟，加拿大和美国（例如CALO， AMIDA， CHIL， ICSI）有多个并行的多伙伴会议分析研究工作。会议分析是国际上语言处理、人机交互、语音研究和视觉处理研究的一个重要领域。这些研究团队在略有不同的数据集上使用不同的技术。我们打算在其中两个大型项目中进行重点的美英联合研究，涉及主要合作伙伴，使我们能够确定工作中有重叠的地方。-我们将通过一个探索性的英美具体项目，共同探索这些潜在的重叠和新的机会，通过评估现有的语料库，确定是否可以使用基于笔的最终用户注释来识别会议中感兴趣的领域。目前确定会议关键部分的技术依赖于语言数据的语义分析，而这种分析尚未产生一致的结果。相反，我们计划探索用户的笔记行为是否提供了关键信息讨论时的可靠指标。Sheffield已经开发了收集这种终端用户注释的工具。我们将与斯坦福大学合作，使用他们的技术和算法来分析相关的注释会议语料库，以验证这一假设。-更重要的是，开展这种集中的合作研究将使我们能够在更大的研究项目中发现重叠之处。这将允许我们产生一个扩展的协作计划，允许更好地利用他人的工作，避免目前广泛的重复。此次访问还将提高英国会议研究在美国重要研究团体中的知名度，因为惠特克已被邀请在加利福尼亚州的许多领先研究机构就这项工作发表演讲。

Following an extended series of meetings, Prof. Stanley Peters of Stanford University, has invited Whittaker and his research team to work with Peters at Stanford in April/June 2008. They will conduct joint research into intelligent extraction of information from recorded meetings. The planned work extends related, but non-overlapping research, already carried out by both groups (Kalnikaite and Whittaker, 2007, Purver et al., 2006, 2007, Tucker and Whittaker, 2006, Whittaker et al., 2007, in press). Aims:- There are currently multiple parallel multipartner research efforts in meetings analysis in the UK, EU, Canada and US (e.g. CALO , AMIDA , CHIL, ICSI). Meetings analysis is a major area for international research in language processing, HCI, speech research and vision processing. These research teams are using different techniques on slightly different datasets. We intend to conduct focused joint US/UK research involving key partners in two of these large projects, allowing us to determine where there are overlaps in effort.- We will explore such potential overlaps and new opportunities collaboratively through an exploratory concrete US/UK project to determine whether pen-based end-user annotations can be used to identify areas of interest in meetings, by evaluating existing corpora. Current techniques for identifying critical parts of meetings rely on semantic analyses of language data, and such analyses have not yet generated consistent results. Instead we plan to explore whether users' note-taking actions offer reliable indicators of when key information is discussed. Sheffield has already developed tools for collecting such end user annotations. We will work with Stanford using their techniques and algorithms to analyse a relevant corpus of annotated meetings to test this hypothesis.- More importantly, carrying out such focused collaborative research will allow us to identify overlaps in our larger research projects. This will allow us to generate an extended collaborative plan allowing better exploitation of others' work, avoiding current extensive duplication. - The visit should also raise the profile of UK meetings research in important US research groups, as Whittaker has been invited to give talks about this work at numerous leading research institutions in California.