Random walk on co‐word network: Ranking terms using structural features

Abstract

ABSTRACT This study proposes a weighted random walk method on co‐word networks to identify important themes of a field using structural features of the networks. The goal is to test whether the weighted random walk method can be used to produce meaningful results on co‐word networks. In addition, we examined the relationships among the results from the random walk method and other two common metrics for identifying important themes in a field: frequency and point centrality. Using a dataset of 17K bibliographic records for the articles in the LIS field from the Web of Science, our results indicate that all three measures are significantly correlated. A detailed comparison of the top terms ranked by the three metrics from the years of 2002–2006 and 2007–2012 is provided. The results show that the three measures are generally similar in revealing hotspots and development of the field. However, some noticeable differences are also found. The random walk method boosted the rankings of some lower ranked terms in the other two metrics (e.g. “universe”, “servic” and “develop”) due to their cooccurrences with top ranked terms (e.g. “information”). The findings of this study help to understand the use of random walk method on co‐word networks.