基于核Batch SOM聚类优化的语义Web服务发现机制研究  

Research on Semantic Web Services Discovery Mechanism Based on Kernel Batch SOM Neural Network

作　　者：陈蕾[1,2] 杨庚[1,2] 张迎周[1,2] 陈燕俐[1] 

机构地区：[1]南京邮电大学计算机学院,南京210003 [2]北京邮电大学网络与交换技术国家重点实验室,北京100876

出　　处：《电子与信息学报》2011年第6期1307-1313,共7页Journal of Electronics & Information Technology

基　　金：国家973计划项目(2011CB302900);国家自然科学基金(60873231;60973046);江苏省自然科学基金(BK2009426);江苏省普通高校研究生科研创新计划项目(CX10B_195Z);北京邮电大学网络与交换技术国家重点实验室开放基金(SKLNST-2008-1-04)资助课题

摘　　要：如何快速、准确和高效地发现满足用户需求的Web服务已成为制约服务发展的瓶颈之一。该文针对现有Web服务发现机制中存在的效率低下和查准率不高的两个主要问题,提出了一个基于核Batch SOM神经网络聚类优化的语义Web服务发现框架。该框架分别在服务表示阶段引入WordNet和隐含语义索引技术对Web服务进行语义扩展和概念语义空间降维;在服务分类阶段利用核机学习理论改进一类适用于Web服务分类的核Batch SOM神经网络算法;在服务匹配阶段提出一种基于核余弦相似性测度的Web服务匹配算法。最后,真实Web服务数据集上的实验结果验证了所提出方法的可行性。With the rapid growth and wide application of Web services, the research on how to accurately, efficiently and rapidly find the desired Web services has become a challenging subject. In order to improve the efficiency and precision for Web service discovery, a semantic Web services discovery framework based on Kernel Batch SOM neural network is proposed. Firstly, by introducing the WordNet and Latent Semantic Index （LSI） into the VSM lexical vectors to extend semantics and reduce the dimension, the resulting VSM semantic vectors can well describe Web services＇ true semantic characterization; Secondly, by using the kernel trick to modify Regular Batch SOM＇s weight updating rule, a kernel Batch SOM neural network is proposed to cluster Web services automatically; Thirdly, a kernel Cosine-based similarity matching mechanism is presented to well estimate the similarity of Web services. Finally, the experiments performed on the real-world Web services collection demonstrate the feasibility of the proposed approaches.

关 键 词：WEB服务发现 自组织映射神经网络 WORDNET 隐含语义索引 相似性度量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]