Bias detection in Philippine political news articles using SentiWordNet and inverse reinforcement model

Abstract

Not all information posted on the internet is deemed ‘trustworthy.’ Some articles, especially those related to politics, seem to display traces of bias, whether they be for or against the Philippine administration. This research aims to determine if a news article—and by extension, a news outlet—is biased based on its sentiments and use of lexica. Data were harvested from chosen websites and news outlets provided by Alexa. These data underwent pre-processing and were scored based on their sentiments with the use of SentiWordNet. The resulting scores were then fed into the Inverse Reinforcement Model, which determined whether an article is biased or not. With the use of Inquirer, Philstar, Manila Bulletin, The Manila Times, and Journal Online news articles, the system was able to detect bias with an accuracy rating of 0.89, precision of 1, recall of 0.60 and F-Measure of 0.75.