STIMULATE: An Environment for Illustrated Briefing and Follow-up Search Over Live Multimedia Information

该项目提出了一种环境，提供有关感兴趣的主题的最新简报，将用户连接到相关多媒体文件的综合集合中。基于用户配置文件或查询，该系统过滤多媒体信息以匹配用户兴趣，并自动生成匹配文档集的摘要，为用户提供相关信息的概述。该系统还提供一组符合用户兴趣的具有代表性的图像或视频，以及指向文档本身的超文本链接。用户可以使用多媒体查询来优化相关图像、视频或其他文章的搜索。该研究包括三个主要部分：摘要、多媒体搜索和操作、事件跟踪。研究结果将提供以下方面的进展：使用语言生成对从多个实时新闻来源提取的信息进行总结，使用视觉特征对压缩图像进行多媒体搜索和分类，以及通过检测文件之间的表面差异来跟踪事件。研究还将产生帮助用户跟踪特定新闻领域的多媒体信息的技术，从而减少最终用户（如记者）必须访问的文件数量。

This project proposes an environment to provide up-to-the-minute briefings on topics of interest, linking the user into an integrated collection of related multimedia documents. On the basis of a user profile or query, the proposed system filters multimedia information to match the user interests and automatically generates a summary of the set of matching documents to provide the user with an overview of the relevant information. The system also provides a representative set of images or video that match user interests as well as hypertext links into the documents themselves. The user can follow up with multimedia queries to refine the search for related images, video, or additional articles. The research has three main components: summarization, multimedia search and manipulation, and event tracking. Results will provide advances in summarization using language generation over information extracted from multiple live sources of news, in multimedia search and classification over compressed images using visual features, and in tracking of events by detecting surface differences between documents. Research will also result in technologies to aid a user in tracking multimedia information in specific news domains, reducing the number of documents end users such as journalists must access.