Abstract

This study seeks to present and validate analytical techniques for evaluating figures involved in political corruption cases using an automated news classifier that combines artificial intelligence, linguistics, and social and political sciences. To this end, a topic analysis using LDA (LatentDirichlet allocation) is analyzed using automatic language processing strategies. A large corpus of information extracted from Sophia Search, an automated news classifier, contains 33,699 news items about seven corruption cases considered scandalous from 64 Chilean online news outlets. From this corpus, the analysis of32 figures involved in seven events considered politically scandalous is obtained. These items were triangulated with 1120 concepts that helped to deepen the evaluation ofthe figures. Thus, the results reveal differences and similarities across one axis between media considered hegemonic and non-hegemonic, and across another axis between media considered traditional or emerging, highlighting distinct strategies in each media type. Finally, a conclusion is drawn from the results and future projections are made regarding the use of artificial intelligence tools in social and political sciences.