The Research Of Ontology Matching Based On Text Classification

Semantic web is an extension of present web and its information are put to the clear meanings, this makes the good cooperation between machine and human being. Ontology is the basis of semantic web. Concepts defined in ontology are used as metadata to tag the web content.Comparability of concept semantic is a hot topic in artificial intelligence area. Similarity model of artificial intelligence hammers at computing concept similar from given knowledge. This dissertation presents the whole framework matched with ontology concept, which based on the text classification of machine learning. An ontology concept similarity model founded on instance is described in virtue of analyzing individual characteristic of improved Bayes classifier and support vector machine. The concept similarity algorithm based on text classification is put forward and the test strategy that concept satisfiability and ontology consistency after concept matching is also introduced. Finally the ontology matching theory is adopted in the conceptual knowledge learning system of network education. This thesis includes following contents:①The relevance basic theory is introduced, including semantic net, ontology, text classification and description logic reasoning.②Text classifier based on Bayes is studied. Naive Bayes classifier is proved to be one of the simple and effective classifier and be used widely and successfully. Its performance has equivalence with or has advantage over other typical classification algorithm. The mutual information is introduced and Bayes classifier based on character relevance is proposed in this thesis. The improved Bayes classifier has higher precision and recall for majority text. In fact, the mapping relation between traditional character and concept, concept and class are set up for introducing semantic characteristic.③Support vector machine based on statistics has the advantage of theory maturity, global optimization, and good adaptability and popularize ability. Binary tree architecture support vector machine for multi-class problems is discussed. An improved strategy of binary tree architecture support vector machine algorithm is proposed. Experimental results verify that improved binary tree architecture support vector machine algorithm has higher classified precision and achieve the prospective purpose.④The ontology concept matching framework based on text classification is presented. The main idea is that the characteristic of text classification is obtained after training the text sample of ontology concept. Through cross learning of text data among ontology concept, the similarity evaluation matrix for all concept pair between ontology is achieved. For taking full advantage of multiple classifier the Highlight strategy of concept pair is described in the concept matching procedure, while the assisted classify strategy using ontology semi-structure information is given for overcoming single classifier is not sensitive for some text. The experimental result shows the ontology concept matching algorithm has better precision.⑤Ontology consistence and satisfaction can be carried out grounded on description logic SHIF(D)and reasoning mechanism. The precondition that realizing ontology test using reasoning is that to complete instance data matching and correlation dispose. This is a big workload for ontology matching evaluation. Ontology matching evaluation technique of semantic oriented is given. This method develops relation of ontology concept and has important references for real engineering application.⑥Network teaching system accelerates the education development; essentially it is the traditional extending. Conceptual intelligence learning system model (CILSM) is presented in this thesis. The CILSM utilizes web resources, organizes them into manageable form for computer reading and deducing according to the resource (knowledge) native attribute and teaching attribute, forms knowledge space. Knowledge space is depicted by RDF/RDFS metadata and structured within ontology described by OWL. Knowledge space involves ontology and ontology matching is a big issue. The ontology matching framework and algorithm introduce in this thesis settles this problem in a certain extent.