Bursty Research Topic Detection From Scholarly Data Using Dynamic Co-Word Networks: A Preliminary Investigation

Discovering emerging research topics from scholarly data is crucial to facilitate the understanding of trends and history of a target field. In a traditional science mapping approach, a co-word network, which shows the co-occurrence relationship between words, is depicted using a set of papers published in each of time periods. To identify bursty research topics in dynamic co-word networks, as a preliminary investigation, this paper proposes to introduce a scheme that identifies sudden increase in frequency of each word pair. Specifically, we first apply a burst detection model to a time-series of edge weights in the networks, and then reconstruct a network that consists of bursty word co-occurrences for each time period. To show the effectiveness of the proposed framework, we present a case study on conference papers in the fields of information retrieval, data mining, and WWW. The results of the experiments demonstrate that our method can reflect the research trend as compared with original co-word networks.