【作者】 黎星星；

【导师】 朱庆生；

【作者基本信息】 重庆大学 ， 计算机软件与理论， 2003， 硕士

【摘要】 万维网正在由网页集合转变为彼此作用的服务集合,Web服务为电子商务的进一步发展(尤其是B2B模式)提供了一种契机。然而随着Web服务的发展也显现出了一些急待解决的问题。例如服务之间互操作性的实现基础之一是基于服务功能的自动搜索推荐能力,但当前基于UDDI和WSDL的Web服务查找技术仅使用一些语法信息而非语义信息,不能满足服务功能匹配计算及推荐的需求。这本质上是一个语义问题,因为我们需要从所提供的服务广告和所请求的服务表示之间提取出语义信息,计算这二者的语义相似度。论文采用了先前已趋成熟的语义网技术和推荐系统技术来匹配服务请求和广告,提出了一个同时以语义、语法相似度计算为中心的Web服务推荐系统模型,设计实现了一个实例系统——ServiceSeek,并进行了实验数据分析。论文的主要研究内容如下:(1)论文对现有Web服务查找技术(UDDI和WSDL)进行了分析研究,分析了其不能应用于服务自动查找的原因,以及现有使用过程中的缺陷;论文重点分析了能使字符具有意义(meaning)的语义网技术,说明了对其引用的原因和意义;论文对推荐系统这种个性化服务技术进行了分类讨论,研究了其不能简单应用于Web服务动态查找及推荐的原因,探讨可为论文所用的一些设计思想。该部分的研究对把握国际上有关的最新技术具有重要意义。(2)论文提出了一个基于语义网技术进行Web服务推荐的新模型,其核心思想是:对Web服务请求者发出的服务查询,先根据输入输出匹配算法构造一个粗选集,再进行语法匹配和语义匹配过程,合并结果集再降序推荐给Web服务请求者。这种模型的优点在于:输入输出匹配算法能首先筛选出输入、输出参数的数量和类型能满足服务请求的粗选集,从而有效降低系统计算量;对于服务名称、服务描述等属性,语义匹配算法能允许用户进行模糊查询,包括错误拼写、不完全拼写等,为请求者的使用带来极大方便;语法匹配能跟据概念之间的语义相似度,挑选出距离请求者输入概念最近的服务广告记录,免除了请求者“文海捞针”、“文不对题”的麻烦。论文模型还就服务质量(Qos)的时间、可靠性和价格三方面提出了一个简便的初评算法,帮助请求者从服务执行角度来选择Web服务。(3)论文作者设计了一个ServiceSeek! Web服务推荐系统,该实例系统采用Websphere Studio Application Developer和DB2平台进行开发,用J2EE技术有效实现了论文提出的推荐模型,。其包括:进行Web服务查询和推荐的ServiceMatch模块;验证概念之间语义相似度的ConceptMatch模块;起基础作用的服务编辑ServiceEdit模块。论文还进行了实验数据分析,验证了模型设计思想的正确性。 <WP=5>论文借鉴了基于内容推荐系统的设计思想,采用语义网技术来挖掘请求者输入和已有Web服务记录之间的联系,并用连续值降序表示出来,克服了现有UDDI基于关键词的简单语法匹配布尔查询的不足,论文设计的语法推荐模块也因具有模糊查询功能而具有优势。该模型的研究,在Web服务智能推荐领域进行了有益的探索,具有一定的研究价值。更多还原

【Abstract】 Along with the appearance of web service, the World Wide Web is developing from a colletion of pages into a colletion of services. It promises a bright future for E-commerce, especially for B2B E-commerce. However, there emerge some unsolving problems with the application of web service acting as obstacles of populization of this promising technology, among that automatically intelligent recommendation of web service should be properly settled. Considering their own capabilites, we can not rely on the UDDI or the WSDL technology to implement the task. As it is essentially a matter of semantics, the process of exacting the semantic information for service requester’s queries and service provider’s advertisements is needed so as to make it possible to compute the similarity between a query and a advertisement. Based on the the technologies of semantic web and recommender system, this paper proposed a new model to recommend web service according to the matchings between usser’s request and provider’s advertisement. A demo system, ServiceSeek!, is developed based which experimental data analysis is processed. The following is the main content of the paper:(1)This paper explored the current web service search technologies such as UDDI and WSDL, investigating their strengths and weaknesses. Then the semantic web technology is introduced which can making the characters computers process meaningful,and it is expounded that why we should utilite it. Moreover, the widely applied technology of recommender system is classified and deeply investigated, discussing how to make use of it’s design ideas.This part of study is important for mastering the last technologies development trend. (2)In this paper, a system architecture for web service recommendation based on semantic web is proposed, whose key idea is: After the system accepts a web service requester’s query, a rough set is constructed by ioMatch module. Then synmatic match and semantic match are independently conducted, whose result sets are combined and the top N records is recommended to the requester. The advantages of this system include: the building of the rough set by ioMatch could decrease the computation task; the synmatic match make it available for requesters to conduct a fuzz query, tolerating requester’s misspelling and incomplete spelling; the semantic matching select the closest service records based on semantic similarity so as to release requesters from heavy workload. What’s more, the paper devise a evaluation algorith to rate the quality <WP=7>of service according to it’s executing time, reliabiligy and price. (3)ServiceSeek! web service recommender system is an application system build by J2EE technology, developed on IBM’s Websphere Studio Application Developer and DB2 platforms. The application includes three main functions: the ServiceMatch accepts query and recommends services; the ConceptMatch verifies semantic similarity between two concepts; and the ServiceEdit editsing web service recording. Based on ServiceSeek, experiment data is collected and analysed.This dessertation exploit the advantages of content-based recommender system and semantic web, enhancing the quality of web service query and providing high efficiency for service requestes. The study in this paper is importance to intelligently recommend requesters with satifying web services.更多还原