Abstract: Regarding to the complex network composed of the vast amount of tags in social tagging systems in Internet with their co-occurrences, the weights as the statistical semantic similarity of tag-tag edges and two abstract operators for weights computation were introduced, and a model of tag semantic similarity measurement is established. Comparing with traditional "users-items-tags" tripartite graph based statistic measures or network topology focused nodes similarity measures, this model provides a well defined formal system, which explicitly addresses both the statistical influential factors and the topological influential factors in computation of tag semantic similarities. A cluster of concrete implementations of the abstract operators are devised, which have similar format with T norms and S norms in fuzzy logics. In this cluster, concrete operators of different types or addressing different scopes of network topological factors are configurated with particular parameters (e.g.,parameter h and order l). By incorporating the AUC index and precision index in link prediction of complex network, an experiment is conducted to analyze the effectiveness and feasibility of these concrete operators. The experimental results show that these concrete operators introduce the effects of "semantic complementation" as well as the effects of "semantic destruction" when they are applied, but lower ordered calculations (e.g., 2≤l≤5 in the model) with these operators are helpful for precise analysis of tag semantic similarities, therefore they are useful in devising high accurate tag-aware recommendation algorithms for social tagging systems.