Community structure is a key feature in graphs, where clusters of nodes exhibit strong internal connections and shared properties. In social network graphs, identifying communities based on users’ sentiments toward a particular topic still remains a challenge for the researchers. In this paper, we are proposing a sentiment orientation-based community detection method where the users in the network are not only socially connected but also express the same opinion regarding a particular topic. This research proposes an unsupervised sentiment analysis approach that incorporates point-wise mutual information (PMI) to find out the sentiment of a user on a particular topic. This opinion inclination values are explored to measure the similarity between users. Finally, this similarity value is applied as the edge weight between the users of the created social network graph. Subsequently, a weighted graph clustering algorithm is used to find densely connected users with the same sentiment regarding a specific topic. Results from experiments conducted on real-world social network data demonstrate that our method is effective, and sentimental information is beneficial in mining community structure.