The Analysis And Evaluation On Relation Closeness In Linked Data

With the development of semantic web,linked data is widely used as a kind of structured data model.The analysis of relation closeness in linked data can help people have an intuitive understanding of the importance of the relation and the relatedness between entities.Relation closeness refers to the importance of relation which is relative to the related entities in linked data.At present,there is no specific work to study the relation closeness in linked data,so the method which utilizes the Wikipedia to analyze the relation closeness in linked data is proposed.The work of the thesis is listed as follow:(1)The method which is based on Wikipedia is proposed to analyze the closeness of relation in linked data.Firstly,the relation in linked data is mapped to the relation in PATTY.PATTY contains relations that are extracted from Wikipedia and the corresponding Wikipedia examples of these relations.Then the corresponding Wikipedia pages can be found through the instances of PATTY.Finally,the features of Wikipedia pages are used to analyze the relation closeness in linked data.(2)The method of relation disambiguation is proposed in the thesis.The method solves the problem of ambiguities while mapping the relations in the linked data to the relations in PATTY.Text similarity,relation popularity and class relatedness constitute the input feature vector.The relation disambiguation problem is treated as a special binary classification problem,and the decision tree is employed for relation disambiguation.(3)Two methods are put forward to analyze the relation closeness,and the two methods are heuristic rule method which is based on the relation position in Wikipedia and RelationRank method which is based on random walk model.The heuristic rule method which is based on the relation position in Wikipedia and this method uses this feature of Wikipedia to model and analyze the relation closeness in linked data.The RelationRank method which is based on random walk model uses random walk model to calculate the importance of relation in the Wikipedia page,therefore the relation closeness in the linked data is analyzed.The method proposed in this thesis is verified on the real linked data,the experiment results show that the proposed method is effective and RelationRank method which is based on random walk model is better than the heuristic rule method which is based on the relation position in Wikipedia.The work of this thesis provides a new way for the study of relation closeness in the linked data,which can help people intuitively understand the relation between entities in the linked data.