Abstract: Ontology matching is one of the main methods to slove the problem of ontology heterogeneous.An efficient and precise similar messure is a pre-requisite of an ontology matching process.The paper affords an improved method to solve the complexity and impreciseness of the traditional similarity in ontology mapping.Firstly，the candidate matching set is established through classifying the ontology so as to reduce the amount of computation.Secondly，an integrated approach of based-instance computation，based-attribute computation，based-structure computation is designed to compute the concept similarity.Lastly，experimental results are given to demonstrate the effectiveness of the matching approach

Key words: ontology matching, concept simlarity, classify, machine learning

摘要： 本体匹配是解决本体异构问题的主要方法之一，一个高效、精确的相似度计算方法是本体匹配的前提条件，针对目前本体匹配时计算复杂以及计算不精确的问题，提出了一种改进的本体匹配方法，该方法首先通过对本体库分类来确定候选匹配集，减少了相似度计算的工作量；进而根据本体的定义模型，从概念实例、概念属性和概念结构等方面来综合计算概念相似度，提高了相似度计算的精确度。实验表明该方法能在较少的时间复杂度上达到较好的匹配效果。

关键词: 本体匹配, 概念相似度, 分类, 机器学习