Research On Non-Taxonomic Relationships Learning Based On Domain Concept Knowledge

As an important representation form of conceptual model and relationships betweenconcepts of knowledge, ontology has been widely used in knowledge engineering, semanticweb and information retrieve fields. Its research has being a hot topic both at abroad andhome. Ontology learning which developing rapidly recently is a major way of (semi-)automatically building ontology. The relationships of domain concept, especiallynon-taxonomic relations in ontology learning are urgent to solve. Now non-taxonomicrelationship learning only relies on corpus information, lacking of the research of potentialinformation in existed domain knowledge and heuristic rules in learning procedure. It leads tomore complex learning process, low accuracy result and can not cover the domain knowledge.To solve the above problems, this paper explores the method of acquiring the existedinternal information in domain knowledge, getting the heuristic information of conceptextracting and relation discovering. It starts from the domain knowledge, depends onlinguistics knowledge and domain concept knowledge, analyses them and acquires theheuristic information, propose s a heuristic process of Non-taxonomic relationship, solvestwo core programs in the process: concept extraction and semantic pattern building. Thespecific work of this paper is as follows:1.Analyze and study the significance of non-taxonomic relationships learning,summarize the research background and current research status;2. Introduce the basic theoretical knowledge of non-taxonomic relationships learning;3.Propose a framework of learning the non-taxonomic relationships with conceptheuristic informationConsider the necessity of joining conceptual knowledge processing in the existingcorpus, and analysis of accessing and representation problem of domain concept knowledge.In text corpus, we start from the existing concept knowledge, gain some heuristic information,preprocess the texts, construct the semantic model automatically, and describe anon-taxonomic relationships discovery algorithm using pattern matching. Overall we proposea heuristic non-taxonomic relationships learning frame that contains statistical theory and semantic model.4. To solve the long term extraction, introduce a concept extraction algorithm withDC-value measureFully use the concept knowledge in framework; analyses the concept and collocation,and get the composition rules in domain concept, describe the design of DC-value algorithmand the reason of long term acquiring，use the concept vector to represent the context of term,then introduce the procedure of concept extraction and pseudo code implementation. Last thespecific experiment shows a better result than the Text2Onto tool.5. Present a new automatic pattern acquisition methodTo solve the pattern acquiring in non-taxonomic relations learning framework, we firstaccord on linguistic analysis, construct a meta-pattern set by the iterative training process,automatic acquire mode set in the field of fine-grained, and then describe the pseudo-code ofthe mode iteration. Manual evaluation and comparison of the final extract non-taxonomicrelations, experiments confirmed that it obtain fine-grained non-taxonomic relations;6. Detail a specific instance of non-taxonomic relations learning frameworkResearch the non-taxonomic relationships learning framework with concept heuristicinformation，propose a specific application--the application of non-taxonomic relationslearning in computer science research. We propose a visual knowledge retrieval system,which facilitate the knowledge engineer and the user’s manual evaluation of rapid relationshipfound to provide users with a better intuitive understanding, to facilitate a comprehensiveunderstanding of domain knowledge.We propose a framework of learning the non-taxonomic relationships with conceptknowledge, which extend the research of non-taxonomic relationship learning and supportfast domain ontology building. Furthmore, we combine it with social networks, and discoverpersonalized services. With deep study of the biomedical knowledge, it helps knowledgeclassification and diagnosis of diseases in biomedical field.