URJC-Team at PoliticEs 2022: Political Ideology Prediction using Linear Classifiers

Different works have demonstrated the relationship between personality traits and political ideology and how they influence our daily lives. The challenge proposed in the IberLEF 2022 Task, PoliticEs, consists of extracting political ideology traits from the text by utilising Natural Language Processing (NLP) techiques. This paper describes the participation of the URJC-Team in such task. In particular, the achievement of extracting political ideology features walks through identifying the gender, the profession, and the political spectrum from a binary and multi-class perspective. In this work, we proposed two Machine Learning models to address the binary and multiclass classification problems, a Linear Support Vector Machine and Logistic Regression. The utilized dataset comprises hundreds of tweets that are cleaned and processed to generate various representations that serve as an input for the system. Between the different proposed subtasks, the proposed classification method has obtained competitive results for the binary ideology classification task, reaching 0.81. The proposal has great room for improvement, and we have planned the following steps for it.