Research On Selection And Combination Approach To Ontology Matchers

The purpose of Semantic Web is to build a web that is able to describe things in a way that computers can understand, and to automatically access and exchange between semantic-aware applications. However, due to the decentralized development of Semantic Web, ontologies and applications based on ontology are increasing explosively with emergence of available ontology resources in different granularity, maturity and perspective. To solve the heterogeneity between ontologies, ontology matching became a key interoperability enabler for the Semantic Web. Moreover, the number and type of existing ontology matching approaches and ontology matchers also reflect the importance of ontology matching, and the solutions of ontology matching have great theoretical and practical value on many areas such as Web Service, P2P, biomedicine, e-commerce, information science, software engineering and geographic information system.Recently, ontology matching has received enough attention by research institutions and researchers, and many approaches and techniques have been proposed for specific application domains. But existing ontology matchers are far unable to meet the needs of emerging Semantic Web and other growing ontology applications. Since it is not practical to develop a new matching approach for new demand, selecting, reusing and adjusting existing matchers gradually became an effective solution. Thus, two key issues, the selection and combination methods to ontology matchers, are studied deeply in this dissertation.Firstly, in order to avoid blind matching, an application requirements oriented ontology matcher selection and evaluation approach is studied based on analyzing the characteristics of existing matchers. For a given matching task, this approach uses the application requirements and user preferences obtained by a user interface to select matchers from matcher library, and to evaluate selected matching by an evaluation model. Through an application instance, this approach is illustrated that matchers can be applied properly to matching task, and it has certain practical value to solve the heterogeneity problem between ontologies.Secondly, to combine ontology matchers, a process model for member matcher selection is studied. In order to reduce the difficulty of selection, the matching task is divided into multiple sub-tasks, and then local selection is performed according to function needs of each sub-task. A local selection algorithm based on diversity measure is studied, it recommend many local matcher combination for global selection. Global selection is modeled as a 0-1 integeral programming problem, which is solved by LINGO mathematics software step by step with user interaction. An illustration is used to show the rationality of global selection process. It provides a necessary process derivation for multiple matcher selection.Thirdly, a similarity aggregation method based on induced ordered weighted averaging operator (IOWA) is proposed to realize the inside integration of results from multiple matchers. It first predicts the confidence of similarity, and then takes it as induce value to assign weight to similarity and sums up all the weighted similarities about a given element pair. It can revise the results of the element pairs which are identified as true matches owing to their few abnormally high similarities. Experiment shows the advantage and flexibility of this method. It considers not only the difference between the performances of matchers, but also the difference on matching quality which a matcher executed on different data.Finally, to merge various mappings resulted from multiple ontology matchers, an ontology mapping negotiation approach based on argumentation framework is proposed. A multiple argument relation argumentation framework (M-VAF) is defined, which has four argument relations such as support, disprove, unite and attack. By judging the successful attack and computing the preferred extension, acceptable arguments, namely final mapping results, can be obtained. Experiment results show that this method can improve the effect of other mapping negotiation approaches, and it provides more beneficial support for mapping negotiation among multiple ontology matchers.Research on selection and combination to ontology matchers are carried out in this dissertation, it has very important significance for selecting suitable matcher to different application environments, taking advantages of member matchers, and improving the matching quality and efficiency.