An improved attributed graph clustering method for discovering expert role in real-world communities

Abstract

Question-answer communities are expert centric large-scale information sharing platforms where experts can be incorporated directly or discovered from the communities to provide support to the users who are looking for expert advice. Discovering an expert is a complex task that requires interpretive or structural analysis of the community. The interpretive analysis incorporates techniques such as content analysis, surveys, and ethnography to capture the behaviours and interactions within groups. The structural analysis uses formal methods like structure analysis and clustering to identify the important roles in the community. Structural analysis is mostly used for analysing online communities. Most of the existing expert discovery methods use structural analysis without graph attributes. In this paper, we proposed a structural analysis approach to discover expert role in a support-needed-community by utilising graph attributes. The proposed method is developed specifically for exploration and to accomplish visualisation requirements. We discovered the expert role by using threaded question-answer relationships among people of different occupations. Experimental results demonstrate that the proposed method is faster and effectively find an expert in real-world online communities as compared to existing popular methods.