I2C at IberLEF-2022 DETESTS Task: Detection of Racist Stereotypes in Spanish Comments Using UnderBagging and Transformers.

This paper presents the systems developed for participation at DETESTS (DETEction and classification of racial STereotypes in Spanish) shared task at IberLEF 2022. The objective of the task is to detect stereotypes in sentences from comments posted in Spanish in response to online news articles related to immigration. The proposed systems are based on pre-trained Spanish language models and the implementation of balancing techniques to tackle the problem of imbalance between comments that contain stereotypes and those that do not. Our systems ranked in the top three positions in subtask 1. An UnderBagging-based classifier reached the first position obtaining an F-score of 0.7042.