In this article, we propose and experimentally assess DiSE-growth, which is a tree-based (pattern-growth) algorithm for mining DIverse Social Entities. Our algorithm makes use of a specialized data structure, called DiSE-tree, for effectively and efficiently representing relevant information on diverse social entities while successfully supporting the mining phase. Diverse entities are popular in a wide spectrum of application scenarios, ranging from linked Web data to Semantic Web and social networks. In all these real-life application scenarios, it has become important to analyze high volumes of valuable linked data and discover those diverse social entities spanning over multiple domains in the entire social network (or some social network analyst-focused portions of the network). Moreover, we also extend our algorithm to handle cases where the analysts interactively change their social network mining parameters (e.g., incrementally expanding or narrowing the analyst-focused portions of social networks in which social network mining is conducted). Furthermore, we complement our analytical contributions by means of an empirical evaluation that clearly shows the benefits of our interactive tree-based mining of diverse social entities