Domain Ontology Concept Instances, Attributes And Attribute Values ​​extracted Study

The ontology concept instance, attribute and attribute value are the important elements of domain ontology knowledge base, so the research to ontology concept instance, attribute and attribute value extraction is a basic research work to automatic or semi-automatic construction of domain ontology knowledge base,the accuracy and abundance extent of which directly affects the application performance of domain ontology knowledge base.At present domestic and foreign relevant researches mainly focus on the extraction of ontology concept instance and attribute, or extraction of the pairs of the concept attribute and attribute value have achieved some progress.But in the process of construction and application of the domain ontology knowledge base,only combining ontology concept instance, attribute and the attribute value can have practical significance. This paper focuses on synchronous extraction to ontology concept instance, attribute and attribute value and has made some work as follows:1. To the features of instance, attribute and attribute value of domain ontology concept, they are synchronously extracted combining the advantages of two kinds of machine learning algorithms both CRFs(Conditional Random Fields, CRFS) and SVM(Support Vector Machines, SVM). Firstly, the above problem is converted to a classification problem which sees concept instance, attribute and attribute value as three types of entity, and CRFS model is adopted to recognize entity. Furthermore the SVM model is used to make sure entity corresponding relation on that basis and finally realizes the synchronized extraction of ontology concept instance, attribute and attribute value. Taking some trials on concept instance, attribute and attribute value on Yunnan tourist attractions for instance, the experiment is done to make that the accuracy rate achieves 84.4% and recall rate is up to 82.7% and the F score is 83.6%, which predicts the method has a favorable feasibility.2. This paper introduces entity recognition to ontology concept instance, attribute and attribute value based on CRFS in detail and analyzes their characteristics and properties themselves to ontology concept instance, attribute and attribute value in free structure text, combining with CRFS to express the ability to features of long distance dependence and overlap and the advantages of better to solve problems of mark (classification) offset to make entity recognition.3. Three types of entity corresponding relation extraction of ontology concept, attribute and attribute value based on support vector machine in detail are introduced and analyzes characteristics of physical entity themselves in three types of entity corresponding relation of ontology concept instance, attribute and attribute value and the entity features appeared in sentence structure, combining with the characteristics, support vector machine is suitable for high dimension space to struct optimal class fication hyperplane, to make sure three types of entity corresponding relationship.4. The synchronous extraction prototype system of ontology concept instance, attribute and attribute value are realized and on this basis make an evaluation to extraction of ontology concept instance, attribute and attribute value.