| Resumen |
Hate speech analysis in texts is important, and the development of models for its detection presents a challenge that demands the consideration of various approaches, particularly methods based on natural language processing. The identification of homophobic terms in songs, as proposed in Track 3 of the HOMO-Mex 2024 shared task, is of interest since these events create new knowledge in the area. This paper proposes the utilization of both traditional machine learning and deep learning algorithms to compare their performance. Among the submitted runs, the team achieved the best results using a Decision Tree with the NNLM embedding, attaining a macro F1 score of 0.482, and with a Bert-like model (BETO), which obtained a macro F1 score of 0.486. This represents a non-significant difference, indicating that there is no substantial distinction in the behavior of the models for this problem, and that further investigation is needed since the overall scores were low. © 2024 Copyright for this paper by its authors. |
