Entity Navigation Methods In The Semantic Web

With the rapid development of the Semantic Web,the large number of Linked Data are published on the Web.These entity-centric structured data can be reused to facili-tate a wide variety of applications,which arouses information navigation requirements towards large-scale Linked Data.This thesis concentrates on entity navigation on the Semantic Web,and aims at tackling some navigation problems such as the disorienta-tion and overload recognition.The contents of the thesis are listed as follows.Traditional entity navigation methods only consider the raw link,which describes the explicit relationship between entities.This thesis presents a link pattern-based method.Link patterns derive from the raw links and describe implicit relationships between entities.On top of link patterns,a hierarchy can be constructed to allow explo-ration of linked entities in a hierarchical multiscale fashion.Furthermore,three metrics(informativeness,conciseness and specificness)are proposed to measure the goodness of link pattern.Finally,a link pattern selection algorithm is proposed to consider the three factors of goodness,overlap and coverage simultaneously.The proposed ap-proach is implemented in a Linked Data browser called SView.The approach is eval-uated in a task-based user study and the experiment results show that the approach provides effective support for entity exploration.Traditional entity navigation methods use the link and class separately,and ignore the inter-relationships between the link and class.This thesis proposes a hierarchical coclustering method to simultaneously group links and entity classes.A measure of intra-link similarity and intra-class similarity are introduced respectively,and incor-porated into co-clustering.Finally,users can choose linked entities by using link and class groups simultaneously.The proposed approach is implemented in a Linked Data browser called CoClus.The approach is evaluated in a task-based user study and the experiment results show that the approach provides effective support for entity explo-ration.Based on the traditional random walk model,this thesis proposes a related entity navigation method.Firstly,a notion of users' browsing scenario is introduced to define the scope of user exploration.Then,an entity class association graph is proposed to describe the entity browsing scenario.Based on class association graph,a random walk model is applied to describe the user's exploration process.Finally,an entity selection algorithm is proposed,which not only considers the relatedness of entity and class,but also considers the diversity of class.The experimental results show the effectiveness of the approach based on a golden standard.