The problem of name sharing is widespread in largescale databases or digital libraries,and it causes many research troubles. We propose a graph module named Attributed Relational Graph to describe the figures and links between entities,then we apply an algorithm named ARGResolution based on Attributed Relational Graph to distinct the entities having the same name. The algorithm analyzes the entities and clusters them according to the similarity measure,and eventually gets a set of clusters that correspond to the real entity respectively. The experiment over real datasets shows that mining the links can improve the quality of name disambiguation and resolve the problem successfully.