Do Media Companies Drive Bias? Using Sentiment Analysis to Measure Media Bias in Newspaper Tweets
Embargo until
Date
2018-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
As natural language processing tools are advancing in their use to study corpuses, it is important to harness these tools to study media content for potential media bias. Existing studies have attempted to measure political sentiment by using lexicons to determine whether political texts have a positive or negative connotation and other studies have looked at media bias by manually classifying media content. Although machine learning models are widely used to simulate human decisions about sentiment, few studies have used automated sentiment analysis on newspaper content to measure media bias. This paper uses an out-of-the-box sentiment analysis model on several newspapers’ tweets from four major media companies during the month leading up to the 2016 presidential election. The sentiment analysis results in a sentiment score for each tweet mentioning Republican or Democratic keywords. Overall, this paper finds that only some newspapers had significant differences in the sentiment scores for Republican and Democratic tweets. Additionally, Republican and Democratic sentiment scores were not significantly different between media companies, showing that the companies that own the newspapers may not be driving biased content of the individual newspapers.
Description
Keywords
sentiment analysis, media bias, natural language processing, Twitter