CAREER: Querying Heterogeneous and Multimedia Information Systems

该项目侧重于即将到来的信息管理系统的几个关键需求：对不同来源的集成访问、数据仓库和多媒体数据。它还将教育和研究融为一体。作为该项目教育部分的一部分，加州大学圣迭戈分校的数据库课程正在重塑，为学生提供实践经验、基本原理和接触上述领域的研究挑战。这项研究在两个方面取得了进展：海量互联网资源的整合和视频数据中语义信息的提取和查询。在数据集成方面，开发了查询处理算法，并将其集成到TSIMIS中介器系统中，利用数据复制知识避免检索冗余/复制的数据，提高了系统的可用性和效率。为了提高局部结果的浏览性能，开发了一种新的基于代价的优化器。优化器被提供关于源数据重叠的概率信息，并制定有效传递部分答案的计划。在视频数据方面，开发了COMIX系统。Comix是由用户使用类似SQL的操作和查询定制的，这些操作和查询帮助Comix提取语义信息，速度比暴力方法快几个数量级。Comix引入了新的查询处理优化，利用了视频数据的顺序性、模糊性和连续性。在网络和数字电视信号将进入每个起居室的即将到来的时代，TSIMIS中介系统的拟议版本将允许用户以一种利用Web数据的冗余和复制的方式搜索海量的Web数据，而不是像目前的情况那样受到它的阻碍。Comix系统将是迈向“智能录像机”的一步，它将筛选数字电视信号，并提取用户感兴趣的信息。Http://www-cse.ucsd.edu/~yannis

The project focuses on several crucial requirements of upcoming information management systems: integrated access to heterogeneous sources, data warehousing, and multimedia data. It also integrates education and research. As part of the educational component of this project, UCSD's database curriculum is being reshaped to provide students with hands-on experience, basic principles, and exposure to research challenges in the above areas. The research enhances the state of the art in two areas: the integration of very large numbers of Internet sources and the extraction and querying of semantic information from video data. In the area of data integration, query processing algorithms are developed and integrated into the TSIMMIS mediator system; they use knowledge of data replication to avoid retrieving redundant/replicated data and enhance the availability and efficiency of the system. A new cost-based optimizer is developed for high performance in browsing of partial results. The optimizer is provided with probabilistic information on source data overlap and develops plans for efficient delivery of partial answers. In the area of video data, the COMIX system is developed. COMIX is customized by users with SQL-like actions and queries that assist COMIX in extracting semantic information orders of magnitude faster than by a brute-force approach. COMIX introduces novel query processing optimizations which exploit the sequentiality, fuzziness and continuity of the video data. In the upcoming era where the Web and the digital TV signal will enter every living room, the proposed version of the TSIMMIS mediator system will allow the user to search the vast volumes of Web data in a manner that exploits redundancy and replication of Web data, rather than being impeded by it (as is currently the case.) Then the COMIX system will be a step towards "smart VCRs", which will be screening the digital TV signal and extracting information of interest to the user. http://www-cse.ucsd.edu/~yannis