A Brief Comparison of Community Detection Algorithms over Semantic Web Data

Community detection is a task responsible for categorizing nodes of a graph into groups that share similar features or properties (e.g. topological structure or node attributes). This is an important task in fields such as social network analysis or pattern recognition that span a large and varied amount of information hiding relations with knowledge. In this sense, an initiative that seeks to extract knowledge from data is the Semantic Web, whose primary goal is to represent Web data into a graph in order to discover facts and relations. In this paper, we developed a strategy to apply community detection algorithms over Semantic Web data graphs. For this purpose, five algorithms were tested to identify groups from a dataset retrieved from the DBpedia knowledge base containing more than 45 thousand nodes and almost 500 thousand edges in the domain of movies. Clustering quality was evaluated by using the modularity measure and the features of the best communities were analyzed.