A Spanish dataset for Targeted Sentiment Analysis of political headlines

JUAN MANUEL PÉREZ; EMILIO RECART ZAPATA; TOMÁS ALVES SALGUEIRO; DAMIÁN FURMAN; PABLO NICOLÁS FERNÁNDEZ LARROSA

SADIO Electronic Journal of Informatic and Operation Research

SOCIEDAD ARGENTINA DE INFORMÁTICA E INVESTIGACIÓN OPERATIVA

Lugar: Buenos Aires; Año: 2023

1514-6774

Subjective texts have been extensively studied due to their potential to influence behaviors. While most research has focused onuser-generated texts in social networks, other types of texts, such asnews headlines expressing opinions on certain topics, can also influencejudgment criteria during political decisions. In this paper, we addressthe task of Targeted Sentiment Analysis for news headlines related tothe 2019 Argentinean Presidential Elections, published by major newsoutlets. To facilitate research in this area, we present a polarity datasetcomprising 1,976 headlines that mention candidates at the target level.Our experiments using state-of-the-art classification algorithms basedon pre-trained language models demonstrate the usefulness of target information for this task. We also provide public access to our data and models to foster further research.