面向服务的知识发现体系结构研究与实现

Research and Implementation of Service Oriented Architecture for Knowledge Discovery

知识发现服务(Knowledge Discovery Service,KDS)作为一种数据、计算、语义密集型的高层服务应用,用户通常需要具备非常全面的知识才能正确使用.如何实现一个面向最终用户的、智能的、有质量保证的 KDS架构面临很多困难.现有的研究提出了利用数据挖掘本体和预测执行时间的方法来帮助用户选择正确并且高质量的 KDS.但是数据挖掘本体只是对数据挖掘的方法进行枚举,无法保证服务的质量,而预测执行时间的方法不能体现KDS本身的特点,因而难以获得满意的服务效果.为了更有效地辅助最终用户在面向服务的体系结构(Service OrientedArchitecture,SOA)上自助地实现知识发现应用,该文提出了一种新的面向服务的知识发现体系结构——SOA4KD,将用户的知识发现需求分为内容需求和质量需求,并提出了扩展的知识发现任务本体 EKDTO,以自然语言的方式进行用户意图获取;在考虑到KDS的服务特性的前提下,充分分析了KDS自身的特点,提出了KDS质量本体KDSQO,采用元学习来进行选择最适合的KDS.相对于目前的体系结构,提出了为最终用户提供高质量知识发现服务的一些新方法和技术,为面向服务的知识发现系统设计与实现提供了一个新的参考模型.

Knowledge Discovery Service (KDS) is a high level and computation, semantic, knowledge intensive application, requires professional domain knowledge to use, there is much difficulty to realize an end-user oriented, intelligent and quality assuring KDS architecture. Current research has proposed data mining ontology and runtime prediction to assist user select correct and high quality service, but data mining ontology only enumerate the method of data mining, fail to ensure the service quality, runtime prediction has not consider the unique characteristic of KDS and the result is always unsatisfying. Aiming at assisting end-user self-build Knowledge Discovery Application on Service Oriented Architecture (SOA) more effectively, this paper proposes a novel Service Oriented Architecture for Knowledge Discovery - SOA4KD. User requirement is divided into content part and quality part. An Extended Knowledge Discovery Task Ontology - EKDTO is proposed. Along with Domain Ontology, it can acquire user requirements through natural language interface. A KDS Quality Ontology - KDSQO is proposed which consider the unique characteristic of KDS as well as characteristic of general service, meta-learning is used to select the most appropriate KDS according to user requirements. Compared with current architecture, this paper proposes some new approaches and techniques for providing end-user high quality KDS, and provides a new reference model to implement SOA for Knowledge Discovery.