A Narrative Framework for Analyzing Partisan Perspectives in Event Discourse
Abstract: Experts from several domains, especially political science, are interested in analyzing political discourse associated with real-world news events. This process would typically require researchers to manually analyze a large collection of news articles on a given event, in order to characterize the underlying partisan perspectives from each side of the political map. Instead, in this work, we propose a systematic approach to summarize partisan perspectives, in an automated manner. Our framework allows us to represent each news article with a predefined structure, comprising of talking points, which we then cluster to identify the repeating themes that collectively shape the narrative of an event. Then, we utilize the resulting clusters to generate a summary for each ideology, left and right, that indicates how each side discusses the event. We show the effectiveness of our framework in capturing partisan perspectives across automated proxy tasks, and human evaluation over a set of events. We release the dataset derived from our narrative framework to the research community.
Paper Type: Long
Research Area: Computational Social Science and Cultural Analytics
Research Area Keywords: Partisan Discourse, Narrative Framework, Political Discourse
Contribution Types: NLP engineering experiment, Data resources, Data analysis
Languages Studied: English
Submission Number: 5682
Loading