Análise de emoções e temas presentes na música popular brasileira com processamento de linguagem natural.

Abstract

This work consisted of analyzing lyrics from Brazilian popular music over the decades, using Natural Language Processing techniques to identify predominant themes and underlying emotions. The importance of this topic lay in understanding the cultural and emotional transformations in Brazilian society over time, as music reflects the emotions, concerns, and experiences of society. To produce the work, 12,542 songs distributed among 1,318 artists were collected and pre-processed, spanning the decades from 1960 to 2020. The main technologies used included the Latent Dirichlet Allocation algorithm for Topic Modeling and the BERTimbau model for emotion classification. The results revealed recurring themes, such as love, and emotional patterns over time, providing valuable insights into cultural and emotional changes in Brazilian society. In summary, the study highlighted the constancy of love as a central theme in the songs, as well as the emotional changes reflected in the lyrics, contributing to a deeper understanding of cultural change in Brazil.