From tweets and newspapers to polls A sentiment study on 2017 United Kingdom general election

Abstract

It has been a heated-disputed topic in the literature whether social media and traditional media could represent public opinion truthfully. Without any doubts, obtaining actual public opinion largely benefits the construction of democratic society. It is therefore of critical importance to determine the accuracy of existing measurements of public opinion. In the current study, we examine whether Twitter posts and newspaper articles – prototypes of social media and traditional media, respectively – measure public opinion accurately. In the context of United Kingdom 2017 general election, we conduct a sentiment analysis on Twitter posts and newspaper articles, and further contrast these results to the results of opinion polls, with correlation analysis and relative differences measures. The results indicate that social media (i.e. Twitter Posts) more precisely reflects public opinion, comparing to traditional media (i.e. newspaper articles). Furthermore, Traditional media can provide a more complex than topics on traditional media. These results demonstrate the application of sentiment analysis on social and political researches and shed lights on future directions of public opinion researches.