Study Of Text-based Chinese Ontology Acquisition

With the research on the semantic web, the innovative generation of internet, the statusof Ontology, the foundation of semantic web, research is boosted gradually. Thus, it is criticalfor us to acquire Chinese Ontology in order to extend and implement semantic web in China.It has become an important subject in the Chinese semantic web to acquire Ontology fromcorpus of different fields, and to enhance the quality of the results for effective application.The paper is focusing on the acquisition of Chinese Ontology with discussion and study.Firstly, we present A Bootstrap Method for Ontology Learning Based on Rules of SentencePatterns in order to solve some problems in Chinese Ontology learning. We introduce theframework of the method, and describle the detailed solution to some key technical problemswithin the framework, such as the acquisition and pre-process of the corpus, the definition ofOntology fragment, and the syntax of the rule of sentence patterns.To enhance the quality of Ontology learning, we bring forward the hypothesis ofOntology isomorphism. We want to make full use of the structural information in existingOntology to guide Ontology learning especially to improve the structural quality. To verify thehypothesis we give a precise definition of the common structure in Ontology called MICISO(Maximum Isomorphic Common Induced Sub-Ontology). Based on the definitions, wepresent the formal proposition of the hypothesis, which is then verifiable with experiment.Based on the hypothesis and the definition we present a novel data mining problem,whose aim is learning ontology to find out the MICISOs and further recommend the commonmeaningful structures. We also provide an algorithm for the problem, and based on them wedevelop a practical tool for mining and checking such structures. With the tool, the algorithmis implemented with quite a few pairs of existing ontologies, and has acquired someinteresting meaningful results, which has preliminarily verified the hypothesis. Besides wehave analysized the factors for the quality of the results.With the analysis of Chinese Ontology learning and improvement of its quality, we havebuilt up a practical tool for Chinese Ontology acquisition. With the tool, we can edit Ontologyconveniently, learn Ontology with sentence pattern automatically, and acquire reusablestructures in existing Ontology. The tool has been embedded in the platform for knowledgemanagement KMSphere, which is providing with necessary Chinese Ontology.Finally, we present the idea of the future work based on the core of the paper, i.e. theBootstrap Method for Ontology Learning Based on Rules of Sentence Patterns, and thehypothesis of Ontology isomorphism.