Evolution of Sexist Themes in Hindi Songs Across Decades
Kritin Kesav 2002- Chitturi · University of North Texas Digital Library (University of North Texas) · 2026
This research investigates the evolution of gender-related language and implicit bias in Hindi film songs across seven decades, from the 1950s to the 2010s. A consolidated dataset of 24,201 songs was constructed by combining movie metadata from IMDb and TMDb, estimating corpus completeness through capture–recapture analysis, and retrieving lyrics from HindiGeetMala and Giitaayan. After extensive cleaning, normalization, and transliteration, the lyrics were analyzed using topic modeling to identify recurring thematic structures in Bollywood music. Latent Dirichlet Allocation (LDA) with coherence-based model selection produced thirteen coherent themes representing major lyrical patterns such as romance, heartbreak, devotion, patriotism, and celebration. To quantify gender associations embedded in the lyrics, this work applied the Single-Category Word Embedding Association Test (SC-WEAT) on FastText embeddings trained separately for each decade and each thematic category. Decade-wise results reveal strong female-associated links to pleasantness and appearance in early decades, with a gradual shift toward more balanced or male-associated patterns in later years. Theme-wise analysis shows that certain lyrical contexts, such as devotion or action-oriented songs, exhibit stronger gender polarization than others. The combined findings demonstrate that Bollywood lyrics encode measurable gender biases that change systematically over time and differ across lyrical themes. This thesis provides a large-scale, data-driven view of how cultural narratives around gender evolve in popular Hindi music and offers a reproducible computational framework for analyzing societal bias in multilingual and creative text corpora.