Statistical topic models are concerned with modeling and extracting latent topics from text data. In many scenarios, there are natural social network and information network structures associated with the text data. How text information and network information can reinforce the inference of topics and communities is a promising problem. We propose a novel solution to this problem, which regularizes a statistical topic model with a harmonic regularizer based on a graph structure in the data. The proposed method combines topic modelingand social network analysis, and leverages the powerof both statistical topic models and discrete regularization.The output of this model can summarize well topics in text,map a topic onto the network, and discover topical communities.With appropriate instantiations of the topic modeland the graph-based regularizer, our model can be applied to a wide range of text mining problems such as author topic analysis, community discovery, and spatial text mining. Empirical experiments on various data sets with different genres show that our approach is effective and outperforms both text-oriented methods and network-oriented methods alone. We will also present follow up effort of this work with applications in social communities.