Community detection is an important and crucial issue in social networks.Using different objects' information can help detect the community structure.However

many existing community detection methods are hardly applied in heterogeneous networks.To address the above problem

we propose a semantic-path based community detection method.This method first calculates the similarity matrix based on semantic paths

obtaining the reliability matrix to build a graph regularization term.Then the nonnegative matrix factorization is employed to achieve the community detection in heterogeneous networks.Simulation on real web data demonstrates that our proposed algorithm can detect the community structure in heterogeneous networks.