Abstract. Community detection in networks is a broad problem with many proposed solutions. Existing methods frequently make use of edge density and node attributes; however, the methods ultimately have dif-ferent denitions of community and build strong assumptions about com-munity features into their models. We propose a new method for commu-nity detection, which estimates both per-community feature distributions (topics) and per-node community membership. Communities are mod-eled as connected subgraphs with nodes sharing similar attributes. Nodes may join multiple communities and share common attributes with each. Communities have an associated probability distribution over attributes and node attributes are modeled as draws from a mixture distribution. We make two basic assumptions about community structure: commu-nities are densely connected and have a small network diameter. These assumptions inform the estimation of community topics and member-ship assignments without being too prescriptive. We present competi-tive results against state-of-the-art methods for nding communities in networks constructed from NSF awards, the DBLP repository, and the Scratch online community. 1