Semantic Macro-Themes in Sertanejo Universitário Lyrics: Topic Modeling with BERTopic

• Amanda Schmieleski Cossa Manfredini UFRGS
• Karin Becker UFRGS

DOI: https://doi.org/10.5753/kdmile.2026.32005

Resumo

Music lyrics provide a rich textual record of affective, social, and cultural themes, whose organization can be explored at scale through natural language processing. This paper investigates the internal thematic organization of 5,037 stanzas extracted from 959 Brazilian country songs released between 2005 and 2025 (sertanejo universitário). The corpus was collected from letras.mus.br and analyzed with BERTopic at the stanza level, allowing thematic variation within individual songs to be captured. Five embedding models were compared as a methodological control, with ptbr-e5-small adopted as the main configuration. The model identified 44 topics, while 2,066 stanzas (41.02%) were classified as outliers, reflecting the challenge of identifying recurrent semantic patterns in short and often figurative poetic texts. Hierarchical consolidation at a 0.90 conceptual cut yielded 21 macro-topics. Human evaluation of the semantic labels assigned to these macro-topics by four annotators validated 18 of them, with a mean score of 4.23 on a five-point scale. The results reveal a predominantly affective-relational thematic structure, centered on passion, romantic suffering, breakup, and reconciliation, articulated with spirituality, drinking, communication, and sociability. These findings show that stanza-level semantic topic modeling can move beyond identifying isolated recurring themes to characterize how affective and social dimensions are organized within sertanejo universitário lyrics.

Palavras-chave: topic modeling, music lyrics, BERTopic, embeddings, natural language processing

Referências

Chen, D., Satish, A., Khanbayov, R., Schuster, C., and Groh, G. Tuning into bias: A computational study of gender bias in song lyrics. In Proc. of the 9th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature. ACL, New Mexico, pp. 117–129, 2025.

Grootendorst, M. BERTopic: Leveraging bert and topic modeling for efficient document clustering. [link], 2022.

Junior, J. S., Rossi, R., and Lobato, F. A lyrics-based approach to knowledge discovery in brazilian music: Sertanejo as a case study. In Anais do XVI Encontro Nacional de Inteligência Artificial e Computacional. pp. 949–960, 2019.

Piceni, H. R., Alexandre, P. V., and Balreira, D. G. A música brasileira na ditadura militar: uma análise de tópicos com bertopic e gsdmm. In Anais do XVI Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana. SBC, Porto Alegre, RS, Brasil, pp. 349–360, 2025.

Requena, B. H. d. A. F. A universidade do sertão: o novo retrato cultural da música sertaneja. M.S. thesis, Universidade de São Paulo, São Paulo, 2016.

Röder, M., Both, A., and Hinneburg, A. Exploring the space of topic coherence measures. In Proceedings of the Eighth ACM International Conference on Web Search and Data Mining. ACM, pp. 399–408, 2015.

Schlösser, A. et al. Representações sociais de término de relacionamentos amorosos em músicas do sertanejo universitário. Psicologia em Revista 22 (2): 407–427, 2016.

Webber, W., Moffat, A., and Zobel, J. A similarity measure for indefinite rankings. ACM Transactions on Information Systems 28 (4): 1–38, 2010.

Yepez Rojas, J. D. and Becker, K. Unsupervised topic modeling of song lyrics with large language models: A zero-shot framework for discourse analysis. Expert Systems with Applications vol. 333, pp. 133782, 2027.