Automated Profiling of the Balance of Optimism and Pessimism in Online News Content

Abstract

Using semantic techniques, we determined a probabilistic score indicating whether news stories were more optimistic (or solutions-oriented), versus their being more pessimistic (or threnodic). We observed over the length of our study that some news outlets, which were comparable in their topical coverage, quantity of output, and geographical focus, differed vastly in their level of optimistic or solutions-oriented news content. This did not seem to correlate with any perceived political bias (left vs. right) nor with the demographic of the target audience, and so raises questions of whether editorial culture or some other causal factor is at work, apart from the typical ideological or audience-driven biases. We found that it is indeed possible on a fully automated basis to profile media sources as falling more on the optimistic or pessimistic side of the spectrum.