Automatic Classification of Online News Headlines
Mark Pope · Carolina Digital Repository (University of North Carolina at Chapel Hill) · 2019
The rise of online news over the past decade has altered how individuals obtain news and this study sought to determine the types of online news headlines most often selected by news websites as their "Top Stories". Headlines from four news websites were downloaded using Really Simple Syndication (RSS) feeds. Supervised learning was conducted with the downloaded headlines to develop models which could automatically classify each website's "Top Story" headlines, whose specific news category was unknown. "Top Story" headlines were also matched to headlines with known news categories from the same period to determine which news categories were most often represented as "Top Stories". The results show that some news categories' headlines, particularly those that had unique terms, were classified correctly based on the text contained in the headline. Headlines from World and US/UK news categories were most often represented as Top Story headlines, followed by Business, Politics, and Entertainment.