Computational Approaches to Detection of Narrative Frames in News

Boon, Miriam

doi:https://doi.org/10.21985/n2-36me-7v07

Work

Computational Approaches to Detection of Narrative Frames in News

Public

Download PDF

Download All Files (.zip)

Citizen media literacy is essential in a democratic society, particularly in the online environment where valid media sources have proliferated alongside purveyors of fake news. This dissertation explores technologies that automatically detect aspects of bias in news articles, with the ultimate aim of leveraging them to augment media literacy. It describes several research projects in this vein, including using a machine learning model to identify election news with a ‘horse race’ frame, using crowdsourcing to identify use of narrative frames such as ‘David vs. Goliath’ in news articles, using machine learning to automatically detect those narrative frames, and detecting entities who are treated as heroes, villains or victims in news stories. This research leverages natural language processing, artificial intelligence, and machine learning methodologies, including information extraction, sentiment analysis, and named entity recognition. The prototype systems developed in this work achieved reasonable performance, with soe interesting caveats in one case. From an NLP perspective these findings reinforce the importance of inspecting the features that arise when applying simple machine learning methods to linguistically complex problems. From a communication studies perspective, the main contribution is to show how narrative framing in news can be studied computationally, and to make the resulting tools available for use in media studies. And from a human-computer interaction perspective, the literature review provides evidence to support the theory that surfacing implicit characteristics of news stories such as narrative frames and narrative roles can mitigate their effect, while our findings provide evidence that such systems are technologically feasible. Thus, user studies to more directly establish the effects of such systems are warranted.

Creator