A system for research community annotation

Abstract

Community detection as an important task that allows understanding of complex networks has attracted much research attention. However, existing research stops at uncovering the community structures, from which the semantic meaning of individual communities cannot be inferred directly. This paper proposes to semantically annotate communities detected in a network. In particular, a system for annotating the research communities in a bibliographic network is developed. An unsupervised method is adopted to generate annotations according to three criteria, 1) the annotations should be semantically relevant with the community, 2) the annotations should be discriminative across communities, and 3) the annotations should have high topic coverage. The system shows that the generated annotations effectively enable the interpretation and understanding of large and various research communities, advancing existing research into community detection for community comprehension.