基于主题模型的Web服务发现方法研究

发布时间：2019-06-28 19:16

【摘要】：Web服务作为一种自主而开放的应用实体,具有松散耦合、平台无关、互操作性强等特点,特别适合在Internet环境中发布和使用。随着互联网上Web服务数量的快速增长,如何从中快速有效地获取满足需求功能的Web服务,这是Web服务发现研究的内容。目前常见的Web服务发现方法主要有基于关键字的方法和基于语义的方法。其中基于关键字的服务发现不能理解用户语义信息,导致查全率不高,传统的基于语义服务发现方法由于限制太多,推广性不强,基于主题模型的Web服务发现是一种特殊的语义发现方法,它相比于传统的语义发现方法具有限制条件少,推广性强等优点。本文分析和研究了 Web服务发现及主题模型等相关技术,对Web服务文档特点、相似度计算方法以及聚类算法进行了深入研究。首先根据WSDL描述文档特点,结合国内外文本数据处理方法,对文档进行特征提取、去除停用词与标签、连接词分词、大小写字母转换、词干还原等,得到数据处理后的文档词汇集合。然后基于BTM对数据处理后的文档词汇集合进行主题建模,使用Gibbs采样对主题进行训练,针对主题个数不确定问题,通过计算主题结构相似度来确定主题数目,得到文档主题信息。接下来对文本相似度度量方法进行了研究,使用服务-主题向量与服务-特征词权重向量线性结合的方式来计算服务之间相似度。对聚类算法进行了研究,为了提高计算效率,使用基于最大距离法的k-means算法对Web服务进行聚类,得到Web服务簇集合。最后,对Web服务进行查询时,找到与查询最相关的Web服务簇,把簇中相关度高的Web服务作为发现的结果。本文实现了方法中的具体模块。最后利用测试集对该方法进行实验验证,实验结果表明,使用本文提出的方法进行Web服务发现具有较高的查准率。并且本文的研究对Web服务发现的相关处理过程具有借鉴意义。
[Abstract]:As an independent and open application entity, Web service has the characteristics of loose coupling, platform independence, strong interoperability and so on. It is especially suitable for publishing and using in Internet environment. With the rapid growth of the number of Web services on the Internet, how to obtain Web services that meet the requirements quickly and effectively is the content of Web service discovery. At present, the common methods of Web service discovery are keyword-based method and semantic-based method. Among them, keyword-based service discovery can not understand user semantic information, resulting in low recall. The traditional semantic service discovery method has too many limitations and low generalization. Web service discovery based on topic model is a special semantic discovery method, which has the advantages of less constraints and strong generalization compared with the traditional semantic discovery method. In this paper, Web service discovery and topic model are analyzed and studied, and the characteristics of Web service document, similarity calculation method and clustering algorithm are deeply studied. Firstly, according to the characteristics of WSDL description document, combined with domestic and foreign text data processing methods, the document feature extraction, removal of deactivated words and tags, conjunction word segmentation, case and lowercase letter conversion, word stem reduction and so on, to obtain the document vocabulary set after data processing. Then, based on BTM, the topic set of document vocabulary after data processing is modeled, and Gibbs sampling is used to train the topic. Aiming at the uncertain number of topics, the topic number is determined by calculating the similarity of topic structure, and the subject information of the document is obtained. Secondly, the measurement method of text similarity is studied, and the similarity between services is calculated by using the linear combination of service-topic vector and service-feature weight vector. The clustering algorithm is studied. In order to improve the computational efficiency, the k-means algorithm based on the maximum distance method is used to cluster the Web services, and the Web service cluster set is obtained. Finally, when querying the Web service, the Web service cluster which is most related to the query is found, and the Web service with high correlation in the cluster is taken as the result of the discovery. In this paper, the specific modules of the method are implemented. Finally, the test set is used to verify the method. The experimental results show that the proposed method has a high precision in Web service discovery. And the research of this paper has reference significance for the related processing process of Web service discovery.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1;TP393.09

【相似文献】