Using classification principles to support semantic integration in information engineering

语义问题是跨资源有效、大规模共享信息的紧迫问题之一，例如那些在互联网上可获得的资源。近年来，语义网的前景受到了相当大的关注。要有效地解决语义问题，需要克服两个挑战：(1)为单一信息源指定语义；(2)提供协调独立和异质信息源提供的语义规范的机制。建议的研究将开发新的、基于理论的机制，用于向信息源提供语义规范和协调独立信息源，并将评估这些方法的有效性。该研究计划将重点放在两个领域。首先，它将开发方法来增强结构化数据源(如传统关系数据库)的语义，方法是利用对实例进行分类所可能产生的推理能力。也就是说，当我们对事物进行分类时，我们可以推断出关于该事物的信息，而不是过去用于分类的信息。这些方法的有效性将使用现有的不同种类的数据来源进行评估。其次，研究将使用分类产生的推理能力来为协作标记机制提供语义层，这些机制最近成为流行的工具，用于在社交书签上下文(例如del.icio.us和Flickr.com)中注释资源。我们将通过评估这些方法在促进在流行的社会书签网站上检索关于资源的信息以及在知识管理应用程序中的有效性来评估这些方法。随着联网资源的可获得性和范围的扩大，对有效方法和工具来管理信息的需求变得更加迫切。本研究旨在通过支持在线资源的有效搜索、过滤和使用，为语义网的信息工程做出贡献。

The problem of semantics is one of the pressing issues in effective, large-scale sharing of information across sources, such as those available over the Internet. The prospect of a "Semantic Web" has received considerable attention in recent years.  Addressing the problem of semantics effectively requires overcoming two challenges: (1) specifying semantics for a single information source; and (2) providing mechanisms for reconciling the semantic specifications provided by independent and heterogeneous sources.  The proposed research will develop new, theory-based mechanisms for providing semantic specifications to information sources and for reconciling independent sources, and will evaluate the effectiveness of these approaches. The research program will focus on two areas. First, it will develop methods to enhance the semantics of structured data sources, such as traditional relational databases, by exploiting the inferential capabilities possible as a result of classifying the instances. That is, when we classify something, we can infer information about the thing beyond what was used to classify it. The effectiveness of these methods will be evaluated using available heterogeneous data sources. Second, the research will use the inferential capabilities resulting from classification to provide a semantic layer to collaborative tagging mechanisms that have recently emerged as popular tools to annotate resources in social bookmarking contexts (e.g., del.icio.us and Flickr.com). We will evaluate these methods by assessing their effectiveness in facilitating the retrieval of information about resources on popular social bookmarking sites, and in Knowledge Management applications.As the availability and scope of networked resources grow, the need for effective methods and tools to manage information is becoming more critical. This research aims to contribute to information engineering for the Semantic Web by supporting the effective search, filtering, and use of online resources.