【作者】 Junaid Ahmad；

【导师】 徐德智；

【作者基本信息】 中南大学 ， 计算机科学与技术， 2010， 硕士

【摘要】 语义Web虽发展缓慢但却稳步发展。本体作为语义Web的核心部分受到了广泛关注并且还需寻找一些新的技术来实现本体的发展。为了使语义Web的研究切实可行,毫无疑问,本体映射和映射的评价是我们最很重要的一个研究方向。当前可行的本体映射技术并未充分考虑本体的语义,并且一般都是采用传统的技术来计算本体之间的相似度,因此匹配结果并没有达到实际预期效果。传统技术使用多策略映射技术和通过给每一种映射策略分配固定的权值的方法来合并映射结果,而事实上语义本体的权值应该通过其语义来决定权值的分配。在本文中我们提出了一种通过上下文语义来设置权值和计算相似度的条件信息量的概念。评价映射结果的算法与本体映射一样重要,不同时期出现了不同的映射评价的技术,其中查全率和查准率是最常用的技术。和传统的映射技术一样,传统的评价技术也没考虑本体的语义。本文提出了纯粹的语义查准率和查全率的一个评价框架,这一框架确保评价方法是纯语义的。我们将把这个框架与一些基本应用的评价方法相结合,将其应用实际操作中。更多还原

【Abstract】 The world of semantic web is progressing slowly but constantly. Ontologies, being the backbone of the semantic web, needs to be focused and new techniques are required to be developed relating to it. Ontology mapping and mapping evaluation are undoubtedly one of the most important operations we need to develop in order to make the semantic web become practical for usage. The current available techniques for ontology mapping do not consider the semantics of the ontologies and use the traditional techniques for finding similarities between the ontologies. As a result the alignments produced are not as close or not as good as they should be. Traditional techniques use multiple strategies and combine the results by assigning fixed weights to each strategy, while in actual the ontologies are semantic in nature so the weights should also be assigned according to their semantics. We propose the concept of conditional information quantity which modifies formulas for computing similarity and set the weights according to the semantic contexts of strategies. Just as mapping is important, evaluating the results of mapping algorithms is also important. There are various techniques or evaluating alignments developed at different times, but precision and recall are the most commonly used techniques. The traditional evaluation techniques, just like traditional mapping techniques, do not consider the semantics of the algorithms. Here, we propose a framework for purely semantic precision and recall. This framework will make sure that the evaluation measures are purely semantic. We will also instantiate the framework with some application oriented evaluation measures.更多还原