“…; [1,[13][14][15][16][17][18] study the tags relevance by textual information, for example [1,13,17,18] calculate the tags concurrence probability and analyze the semantical relevance between tags, [14] analyzes the ontological relationship for tags, [1,15] assume that the important tags' positions are prior to those irrelevant ones; [19,20] argue that tags are content-related to the image, and after comparing to other 3 factors, [19] concludes that the image content is the key factor for tag relevance learning.…”