Beats of Bias: Analyzing Lyrics with Topic Modeling and Gender Bias Measurements

2024-09-24

arXiv.org

AI Summary

This study investigates the diachronic evolution and thematic heterogeneity of gender bias in English song lyrics. Methodologically, it analyzes 537,000 English song lyrics using a novel integration of BERTopic for thematic modeling and SC-WEAT (Single-Category Word Embedding Association Test) for quantifying implicit gendered semantic associations, leveraging BERT-based contextualized embeddings and temporal analysis. Results reveal a longitudinal thematic shift from romance toward sexualization of women; words denoting intelligence and strength exhibit significant male association, while appearance- and vulnerability-related terms show strong female association—both biases varying markedly across themes and genres. Additionally, high-density clusters of misogynistic and vulgar content are identified. The work contributes a reproducible methodological framework and a large-scale empirical benchmark for studying cultural bias in music, advancing computational social science approaches to linguistic stereotyping.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Application Domains: Humanities & Computational Social Science

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This paper uses topic modeling and bias measurement techniques to analyze and determine gender bias in English song lyrics. We utilize BERTopic to cluster 537,553 English songs into distinct topics and chart their development over time. Our analysis shows the thematic shift in song lyrics over the years, from themes of romance to the increasing sexualization of women in songs. We observe large amounts of profanity and misogynistic lyrics on various topics, especially in the overall biggest cluster. Furthermore, to analyze gender bias across topics and genres, we employ the Single Category Word Embedding Association Test (SC-WEAT) to compute bias scores for the word embeddings trained on the most popular topics as well as for each genre. We find that words related to intelligence and strength tend to show a male bias across genres, as opposed to appearance and weakness words, which are more female-biased; however, a closer look also reveals differences in biases across topics.

Problem

• Analyzes gender bias in English song lyrics using topic modeling.
• Measures thematic shifts and misogynistic content over time.
• Quantifies gender bias across topics and genres using SC-WEAT.

Innovation

• Used BERTopic for clustering song lyrics
• Applied SC-WEAT for gender bias analysis
• Analyzed thematic evolution and bias scores