Abstract

This study proposes a quantitative method to assess the pertinence of political language on national issues, addressing the complexity of analyzing political discourse and its relevance to citizens’ concerns. Using word embeddings and linguistic models trained on Wikipedia, a “pertinence score” was developed to measure the relevance of political discourse in contexts such as the economy and health. The method was applied to the 2018 Colombian presidential election, revealing significant differences in thematic pertinence between candidates. Survey validation confirmed the correlation between automatic and human scores, highlighting the model’s ability to discriminate ideological positions through lexical analysis.