Measuring the Influence of Mass Media on Opinion Segregation through Twitter

During the US presidential elections the media played a major role in presenting the candidates’ vision on several topics. Nevertheless, the diversity of opinions along with the political currents, one might notice segregation in opinions among some topics related to each other or a candidate. In the meanwhile, posting opinions on social media could be represented as a sentiment vector towards multiple issues. This gives us a ripe ground for clustering opinions to view tweets that hold similar opinions. In this paper we investigate the media's influence on segregating opinions by constructing an aspect-based opinion mining framework. Our main task is to detect the segregated groups of opinions by solving the proposed model using expectation maximization (EM) algorithm. We examined a corpus of tweets collected, which are related to famous political topics. We show interesting observations on the sentiment used for particular topics among the groups of opinions, and conclude the percentages of media influences among the segregated groups of opinions with respect to these topics.