Research On Ontology Automatic Construction Based On Relational Database

Ontology plays a very important role in the Semantic Web.It enables the exchange,reuse and sharing among information based on semantics,thus connecting information into a huge knowledge network.However,most of the current domain ontologies need to be built and edited manually by experts,and due to the information sources are mostly unstructured or semi-structured,the scale of data is large and the quality of them is not very good,the efficiency and quality of ontology construction are not very satisfied.Therefore,how to construct ontology automatically is an important research work in the Semantic Web.The goal of this dissertation is to realize the automatic ontology construction extracting semantic information from the data sources and realize the ontology oncept extension from different ontologies.then it can provide semantic services for more users.The main research work is as follows:Based on relational database,an automatic construction method of OWL ontology is proposed,which solves the problem that most domain ontologies need to be manually built and edited by experts.This method mainly analyzes the relational schema of database and the instances in the tables.By analyzing different types of tables in the database,an algorithm of conversion from relational database to OWL ontology is given.Finally,it realizes the automatic construction of OWL ontology.For the constructed ontology,this dissertation improves the calculation of word similarity in English and Chinese.Then,based on the word similarity calculation,an ontology concept extension method is proposed which can help to extend and integrate the concepts among multiple ontologies.Based on the ontology,we design and implement an ontology semantic retrieval system to provide an engineering design that is more versatile and extensible for ontology automation construction,extension and ontology semantic retrieval.Experiments show that in an e-commerce database,the ontology automatic construction proposed in this dissertation can extract more classes,attributes and axioms of OWL ontology than the open source tool.Our method increases the total number of ontology elements by 87%and has richer semantic information.In English,this dissertation improves the information content model based on WordNet and the word similarity calculation.In Chinese,it improves the structure of CiLin and expand the word library with HowNet.Then,a similarity calculation strategy based on the two Chinese word libraries is proposed.The English and Chinese word similarity calculation can get higher Pearson correlation coefficient compared with other methods on the multiple standard data sets.On average,the Engslish and Chinese increase by 7.4%and 6.6%respectively.The ontology concept extension method proposed in this dissertation gets a weighted similarity using ontology data attributes and object attributes.Then,it can find the most similar ontology concept to be added to extend the existing ontology concepts in an equivalent or synonymous relationship.From above,we can realize the discovery and extension of similar concepts in different ontologies.The ontology semantic retrieval system implemented in this dissertation mainly includes automatic ontology construction and extension module,semantic analysis module and semantic query module.The system has a good scalability and interactivity,it can understand and retrieve the problems input by users and returns a visualization of results.