Dynamic Similarity Search on Semistractured Multimedia Databases Using Multiple Similarity Measures

由于最近互联网和计算机技术的发展，高级信息使用的重要性已经增加。特别是，对于半结构化的多媒体数据，如多媒体XML数据，我们需要技术来实现灵活的信息检索使用多种相似性的措施，如基于数据内容相似性，结构相似性，相对和动态相似性，并结合上下文信息。在这个项目中，我们已经研究了相关的技术问题，实现动态和自适应检索，集成和转换的半结构化多媒体数据，使用这种不同的相似性措施。主要研究内容如下：1.开发了一个可扩展的XML查询处理系统X-QL，它可以在XML数据集的检索、摘要生成和转换中使用不同的相似性度量。在此基础上开发了一个Web信息与应用集成系统.研究了基于关系数据库的XML视图中包含外部函数的查询处理.开发了一个基于实例的XML操作框架AQUA并进行了实验评估.开发了一个集成机器学习技术和XML处理的XML查询学习系统XLearner，并进行了实验评估.以分类法资讯为背景之网路搜寻方法研究。该研究包括使用ODP和Google等真实的网站进行的实验评估。发展一个连续查询处理系统，为移动目标提供邻域信息。移动对象的位置和运动信息被用作信息提取的上下文。研究集成多个流的体系结构和查询处理方案，作为半结构化数据流灵活操作的基础。

Due to recent development of internet and computer technologies, importance of advanced information use has been increasing. In particular, for semistructured multimedia data such as multimedia XML data, we need technology to realize flexible information retrieval using multiple similarity measures like those based on data contents similarity, structural similarity, and relative and dynamic similarity incorporating context information. In this project, we have studied technical issues related to implementation of dynamic and adaptive retrieval, integration, and transformation of semistructured multimedia data using such different similarity measures. The main research items are as follows.1. Development of an extensible XML query processing system named X-QL which enables the use of different similarity measures in retrieval, summary generation, and transformation on XML data sets. A web information and application integration system lias also been developed on top of it.2. Study on processing of queries including foreign functions on XML views built on relational databases.3. Development of an example-based XML manipulation framework named AQUA and its experimental evaluation.4. Development of an XML query learning system XLearner integrating machine learning techniques and XML processing and its experimental evaluation.5. Study on web search methods using taxonomy information as contexts for web search. The study includes experimental evaluation using real web sites such as ODP and Google.6. Development of a continuous query processing system which provides neighborhood information for moving objects. Position and movement information of moving objects are used as contexts for information extraction.7. Study on an architecture and query processing schemes for integrating multiple streams as a basis for flexible manipulation of semistructured data streams.