Session: 5

PS5-18 | Sentiment analysis in news media headlines in 2019 Presidential Elections: Exploratory Reliability Study Analysis

Emilio Recart


Social networks offer the facility to access political news instantly. In recent years, given the spread of fake news, the role of news transmission in the political perception of citizens has been elucidated. Based on this, new computational models have been developed that seek to identify and predict subjective perception towards social content. Reliability has been a widely used tool in measuring agreements that different people arrive at the presentation of the same stimulus. Therefore, sentiment analysis could be a useful tool for analyzing electoral behavior. The purpose of this study was to evaluate the subjective perception of individuals for each presidential formula/force of principal Argentina´s newspapers during 2019 elections using reliability coefficient agreement. For this, 3 participants were recruited to classify 2257 headlines of the principal country’s newspapers as positive, neutral, or negative according to their perception. To minimize ideological bias, each formula/force was
replaced by a “Target”. Krippendorf nominal reliability alpha metric yielded adequate inter agreements between the participants. With this tool, we found that Fernandez was mentioned with a positive connotation in 1136 headlines(negative: 721), while Macri in 932
ones (negative: 1271). Other candidates did not exceed 420 positive mentions(negative<21). According to this, using these metrics could be a useful tool for future studies for classifying the valence of the headlines.