SENTIMENT ANALYSIS OF NEWS ARTICLES AND ITS COMMENTS: A NATURAL LANGUAGE PROCESSING APPLICATION

Abstract

In this paper I present the use of the VADER (Valence Aware Dictionary and sEntiment Reasoner) model in conducting sentiment analysis of news articles. I explore the sentiments of articles in The Global and Mail and comments made by readers. The dataset contains over 10,000 articles from 2013 to 2016 as well as over 660,000 comments. I find that during this period, the sentiments stay rather consistent and average around a score of 0 (neutral), while the sentiments of the comments follow a bimodal distribution, with around 10% of users consistently giving highly positive comments and another 10% giving highly negative ones.