Muses or Stereotypes? Identifying Historical Patterns of Sexism in a Corpus of Brazilian Lyrics

Abstract

This study aims to identify gender bias in Brazilian songs by analyzing the most frequent predicatives used to describe women and the most referenced occupations associated with the feminine gender. To achieve this, we created a corpus containing 146,612 song lyrics and applied natural language processing techniques to extract sentences that describe women. The identified predicatives were annotated and used to train a machine learning model that categorizes them into five descriptive categories. Additionally, we compiled a list of occupations mentioned in the lyrics. From a distant reading perspective, the results reveal a persistent historical pattern of sexism: women are predominantly portrayed through physical and emotional traits, while men are more frequently associated with character and social roles. In the professional domain, women are mainly depicted in caregiving and entertainment roles. These findings align with previous studies and contribute to methodological advancements in identifying gender bias in Portuguese-language texts. The text contains potentially harmful and offensive examples.