【作者】 毛郁欣；

【导师】 吴朝晖；

【作者基本信息】 浙江大学 ， 计算机科学与技术， 2008， 博士

【摘要】 本体作为语义Web的知识表示基础,在构建基于语义的系统或应用中发挥着至关重要的作用。随着本体规模的增长,系统处理和利用本体的效率会降低。对于大规模的领域本体,语义Web应用通常只需要利用其中的部分内容。从目前语义Web和本体的研究来看,还缺乏比较有效的模型与方法,来支持在语义Web应用中对大规模本体的重用。为了构建和推广面向语义Web的应用,有效地管理和利用已有的大规模本体已经成为一个十分现实和迫切的需求。基于上述背景,本文着重探讨了面向大规模本体重用的子本体模型,主要研究内容和贡献包括以下几个方面:□针对语义Web应用在利用本体时存在的局部性,提出了子本体的表示方法。将来自于大规模本体的上下文相关的模块表示为子本体,给出了子本体的形式化表示,并定义了针对子本体的对象操作。语义Web应用能够根据需要动态地抽取子本体,创建特定的子本体知识库。将缓存机制与本体重用相结合,利用子本体缓存作为系统的局部知识库,支持对大规模本体的动态重用。□对面向子本体的推理问题进行了研究。提出了子本体中的基本推理任务,通过特定的子本体推理算法,将本体的推理问题转化为子本体的推理问题,能从一定程度上降低推理的复杂性、提高推理的效率。给出了基于子本体表示的Tableau算法,支持模块化的本体推理。证明了面向子本体的Tableau算法相对源本体而言是半判定的,并给出了保持一致性的扩展推理算法。□针对子本体知识库的优化问题,提出了基于遗传算法的优化方法。该方法对传统的遗传算法进行扩展,提出了基于语义的遗传算法SemGA,使用基于三元组的非二进制编码方式将子本体表示为染色体,根据语义关系执行遗传算子。利用SemGA进行动态地演化,从而达到优化知识库的语义结构的目的。与一般的缓存策略相比,基于演化的方法在效率上和性能上都有比较明显的优势。□面向分布式的Web资源,提出了基于子本体的资源集成与管理方法。该方法利用本体语义对分布式的Web资源进行集成,通过在资源模式与本体之间建立语义映射,实现以子本体为单位的资源管理。将资源匹配过程转化为资源请求与子本体之间的概念匹配,利用遗传算法进行资源优化,满足动态变化的资源请求。模拟实验的结果表明,该算法能进一步提高资源匹配和重用的效率。基于上述工作,同时还设计并实现了一个子本体原型系统DartOnto,支持面向中医药领域的大规模本体重用。通过实例进一步说明了如何应用子本体模型创建中医药知识服务,解决大规模领域本体的重用问题。更多还原

【Abstract】 As the knowledge representation foundation of the Semantic Web, ontologies play a critical role in building a large variety of semantic-based systems or applications. With the growth of ontologies, it will decrease the efficiency of systems in manipulating and using ontologies. However, a specific Semantic Web application often needs portions of a large-scale domain ontology. Considering the existing work about the Semantic Web and ontology, it still lacks of efficient models and methods to support reusing large-scale ontologies in Semantic Web applications. In order to construct and popularize applications towards the Semantic Web, how to manage and utilize large-scale ontologies has become a practical and urgent requirement.Under this background, we mainly talk about a sub-ontology model for large-scale ontology reuse in this thesis. The major research efforts and contributions are as follows:□Considering the locality of Semantic Web applications in using ontology, we propose the representation of sub-ontology. Context-specific portions from large-scale ontology are represented as sub-ontologies. This thesis gives a formal definition of sub-ontology and defines a collection of object manipulations for sub-ontology. Semantic Web applications can extract sub-ontologies dynamically according to requirements and form specific sub-ontology knowledge-bases. This thesis also combines the caching mechanism with ontology reuse to form a sub-ontology cache, which is used as the local knowledge-base of semantic-based systems to support dynamic ontology reuse.□This thesis also presents the research about the sub-ontology reasoning and illustrates the basic reasoning tasks of sub-ontology. The reasoning algorithm based on sub-ontology is used to reduce the reasoning problem for ontology into the one for sub-ontology. In this way, it improves the efficiency of reasoning by decreasing the complexity. A tableau algorithm based on sub-ontology representation is given to support modularized ontology reasoning. We also prove that the tableau algorithm for sub-ontology is semi-deterministic compared with that of ontology. An expansion reasoning algorithm is given for preserving consistency.□Considering the problem of sub-ontology knowledge-base optimization, this thesis presents an optimization approach based on genetic algorithm. The approach extends the canonical genetic algorithm to form a semantic-based genetic algorithm called SemGA. The algorithm uses triple-based non-binary encoding to represent sub-ontologies as chromosomes and performs genetic operators based on semantic relations. SemGA can evolve sub-ontology cache dynamically to optimize the semantic structure of knowledge-base. Compared with the traditional cache policies, the evolution-based approach has benefits in both efficiency and performance.□This thesis presents a sub-ontology based method for integrating and managing distributed Web resources. The method makes use of ontology semantics to integrate distributed Web resources. It manages resources in terms of sub-ontology by creating semantic mappings between the schemata of resources and ontology. The process of resource matching is transformed to the matching between resource requests and sub-ontologies. A genetic algorithm is used to achieve resource optimization to satisfy dynamic resource requirements. The result of simulation experiment illustrates that the algorithm improves the efficiency of resource matching and reuse.On the basis of the work before-mentioned, this thesis also presents a prototype system for sub-ontology called DartOnto. The system is used to support reusing large-scale ontology in the field of traditional Chinese medicine. We illustrate how to use the sub-ontology model in constructing traditional Chinese medicine knowledge service and solving the problem of reusing large-scale domain ontology through a use case.更多还原