Research And Implementation Of Ontology-based Semantic Computation Of Web Page

With the rapid development of Internet, the explosive growth of documents in digital form makes it increasingly difficult for people to obtain useful information and services. Semantic Web which is regarded as the next generation Internet can formally describe the concepts and inference rules, in this way, the traditional World Wild Web takes an important step forward in transferring from information understanding of human-oriented to information understanding of machine-oriented.Ontology is a key technology for constructing Semantic Web, it can normalize concepts and terminologies within a certain domain, so as to help machines to understand natural language. In addition, the semantic computation is the very core of solving semantic heterogeneity problem and achieving semantic sharing and interconnection. Therefore, we do research on Ontology-based semantic computation, aiming to calculate the correlation between concepts through specific knowledge representation.After studying the existing semantic computation especially the LMO algorithm in FALCON-AO system, this paper proposes a new calculation method called OSSC (Ontology-based Semantic Similarity Computation). Virtual documents, instances, attributes, and structure information are considered comprehensively in OSSC, moreover, HowNet is introduced into OSSC as a reference. We hope that the semantic similarity between concepts should be considered in the comprehensive perspective.According to the study, OSSC is implemented in Concept Extract System and applied to subject classification and similarity calculation. OSSC-based subject classification improve the traditional classification in synonymous problem, it can represent the data in a focused way by merging the similar characteristics, there by, the impact of separate characteristics on classification results could be reduced. OSSC-based similarity calculation expresses the similarity between web pages in the form of similarity between concepts, it pays more attention to the semantic information of web pages.Finally, we use standard data sets to test the classification results and the results are evaluated by precision, recall and F1. The experimental results indicate that OSSC can reveal the semantic relationship between concepts and is an effective method to improve the classification results.