Research On Relation Detection And Recognition System Based Ontology In The Field Of Electronic

With the rapid development of Internet and information industry, more and more information appears. Many of the emerging information processing technologies came into being. As the basic research, Relation Detection and Recognition (RDR) became one of the Hot. RDR has wide application in information retrieval, expert systems, vertical search and so on, on which the research has far-reaching significance.This paper analyzes and compares the various techniques of RDR and focuses on the Support Vector Machine (SVM) in the machine learning methods. This article found that the traditional RDR techniques, such as SVM etc, work well on the simple relation extraction. But because these techniques do not have a function of semantic recognition, the extraction of complex relations is difficult. Through researching, this paper found that ontology in artificial intelligence area can effectively solve this problem. So this article has been systematically studied the conception of ontology and deep analysis the ontology modeling elements and construction methods. Combining the features of electronic products this article presents a new domain ontology construction method for RDR system. Then select notebook computers, one type of electronic products, as a representative to construct ontology. Combining between SVM and ontology, this article proposes a RDR method based on merging model. This model uses SVM to extract a single-relation and then utilizes ontology to troubleshoot and merge single-relation in order to form a complex network of relations. This model inherits the advantages of SVM and ontology, the body to solve the problem of complex-relation extraction.Finally, implements the system according to the proposed model. As the special nature of the selected ontology area, paper selects the self-organization corpus to verify the system and obtains a good precision and recall, after analyzed and adjusted.