Understanding the link between audience engagement metrics and the perceived quality of online news using machine learning

Abstract

This article aims to explain the perceived quality of online news articles. Discovering which elements of a news story influence readers’ perceptions could drive online popularity, which is the paramount factor of digital news readership. This work explores an approach to use tree-based machine learning algorithms to address this problem based on selected characteristics, which measure engagement, drawn from prior research mostly developed by communication scientists. A proposed extended model is used to examine the association between the engagement features and perceived quality concerning all the articles depending mainly on their genre. To demonstrate the capacity of using predictive analytics to facilitate journalistic news writing the proposed methodology is applied on a novel data set with 200K articles in total constructed from a blog site. The results of phase A, indicate interesting correlations between the features and the perceived quality of the articles. In stage B, the paper seeks to extract a set of rules that can be used as guidelines for authors in the writing of their next articles, indicating the probability of popularity that their articles may gain if these rules are taken into consideration.