Aim. Having pointed out what we believe to be the deficiencies of past algorithms[4～7], we propose what we believe to be a more efficient one. Section 1 of the full paper briefs how to input our algorithm. Section 2 explains in some detail the principles of our algorithm. Subsections 3.1 and 3.2 explain in some detail the implementation of our algorithm. Its core consists of two phases: (1) during the first phase of compound name matching, our algorithm matches the schemas to be input by running the three linguistic individual matchers of character string matcher, name semantic matcher and domain specific matcher to obtain the name semantic correlation as shown in Fig. 1 of the full paper; (2) during the second phase of feature classification, our algorithm inputs the names and other constraints of schema elements into self organizing mapping neural networks, thus achieving their clustering.%发现数据之间的语义相关性是实现企业信息集成的前提.为克服单一匹配算法难于适应复杂应用环境,提高语义发现的自动化程度,文章提出一种基于混合模式匹配策略的企业信息语义相关性发现方法.在该方法中,定义关系模式作为企业数据源的统一描述和输入,首先通过复合名称匹配过程,计算特征元素名称之间的相似度,然后将匹配结果输入基于神经网络的特征综合信息分类器,实现特征元素的聚类.
展开▼
