Visualizing narrative patterns in online news media

Abstract

News media play an important role in shaping social reality, and their multimedia narrative content, in particular, can have widespread repercussions in the public's perception of past and present phenomena. Being able to visually track changes in media coverage over time could offer the potential for aiding social change, as well as furthering accountability in journalism. In this paper, we explore how visualizations could be used to examine differences in online media narrative patterns over time and across publications. While there are existing means of visualizing such narrative patterns over time, few address the aspect of co-occurrence of variables in media content. Comparing co-occurrences of variables chronologically can be more useful in identifying patterns and possible biases in media coverage than simply counting the individual occurrences of those variables independently. Here, we present a visualization, called time-sets, which has been designed to support temporal comparisons of such co-occurrences. We also describe an interactive prototype tool we have developed based on time-sets for analysis of multimedia news datasets, using an illustrative case study of news articles published on three online sources over several years. We then report on a user study we have conducted to evaluate the time-sets visualization, and discuss its findings.