Social tagging systems is a system that shares online contents and gives a user arbitrary character string (i.e., tag). Many Web services such as Flickr, Twitter, Instagram, Facebook are adopting social tagging system. It is reported that the process of tag growth can be roughly approximated by the Yule–Simon process. On the other hand, the growth of individual tags has also been analyzed. If it follows Yule–Simon process, the probability distribution of deviation can be obtained analytically. However, it is not well studied how the growth of individual tags in the actual social tagging system deviates from the prediction value of the Yule–Simon process. In this paper, we analyze the growth of individual tags of the social tagging system focusing on deviation from Yule–Simon process. Moreover, we propose a model that modifies the Yule–Simon process and explore a more detailed mechanism of tag generation and selection.