Editorial Tag Endogeneity for News Websites

Abstract

Editors and journalists at some news websites label their articles with structure and content-related editorial tags. Each article can have more than one tag and each tag can be used in more than one article. A network of tags can be defined whose edges are all possible pairs of tags in each article. Because editorial tags relate to structure and content rather than individual articles, the analysis of a network of editorial tags could assist editorial decisions to prioritize types of content and articles. In this paper we analyze the network of editorial tags of one of the fastest growing news websites in Portugal, with over 6.1 million visits, 7.6 million page views, and over 1200 editorial tags in 15 months. Standard network characterization reveals a 15.5 average node degree, a 0.794 average clustering coefficient, and a 2.36 average path length, which are indicators of small world and triadic closure effects. We use this tag network to propose endogenous and exogenous models that predict transitions between tags of consecutive article views. The editor can use this tag transition model to prioritize types of articles: articles with endogenous tags to try to promote the reading of articles with similar content, and articles with exogenous tags to try to promote the reading of articles with different content.