Metadata Matters: Identifying Research Communities and Estimating Author Influence in Networks

This work illustrates how clustering of author communities can be improved by using metadata about the authors (such as subject area or institutional affiliation) to supplement network structure based on co-authorship or citation relationships. Using such metadata can improve clustering significantly, both in terms of human interpretability and by allowing detection of smaller groups – but weighting metadata too highly in relation to network structure can be detrimental. The author communities uncovered may subsequently be used to better elucidate the effect of each author on their peers, and so help distinguish between influence and popularity.