Profissões e estereótipos: avaliando o viés de gênero em versões do BERTimbau
Resumo
Este estudo investiga a presença de viés de gênero nos embeddings contextuais do modelo BERTimbau (versões BASE e LARGE), com foco em profissões estereotipicamente associadas a homens e mulheres. Os resultados, obtidos através da métrica WEAT e testes de analogia, indicam um forte viés de gênero em ambos os modelos. Os achados apontam para o risco de que esses modelos reforcem estereótipos em aplicações reais, destacando a necessidade de métodos de mitigação.Referências
Bolukbasi, T., Chang, K.-W., Zou, J., Saligrama, V., and Kalai, A. (2016). Man is to computer programmer as woman is to homemaker? debiasing word embeddings.
Caliskan, A., Bryson, J., and Narayanan, A. (2017). Semantics derived automatically from language corpora contain human-like biases.
Conneau, A., Khandelwal, K., Goyal, N., Chaudhary, V., Wenzek, G., Guzmán, F., Grave, E., Ott, M., Zettlemoyer, L., and Stoyanov, V. (2020). Unsupervised cross-lingual representation learning at scale.
Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Burstein, J., Doran, C., and Solorio, T., editors, Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Vol. 1, pages 4171–4186, Minneapolis, Minnesota. ACL.
Jentzsch, S. and Turan, C. (2022). Gender bias in bert - measuring and analysing biases through sentiment rating in a realistic downstream classification task.
May, C., Wang, A., Bordia, S., Bowman, S. R., and Rudinger, R. (2019). On measuring social biases in sentence encoders.
Puttick, A., Rankwiler, L., Ikae, C., and Kurpicz-Briki, M. (2024). The bias detection framework: Bias detection in word embeddings and language models for european languages.
Santana, B. S., Woloszyn, V., and Wives, L. K. (2018). Is there gender bias and stereotype in portuguese word embeddings? Souza, F., Nogueira, R., and Lotufo, R. (2020). BERTimbau: Pretrained BERT Models for Brazilian Portuguese.
Taso, F., Reis, V., and Martinez, F. (2023). Sexismo no brasil: análise de um word embedding por meio de testes baseados em associação implícita.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. (2023). Attention is all you need.
Yarrabelly, N., Damodaran, V., and Su, F.-G. (2024). Mitigating gender bias in contextual word embeddings.
Yucong Duan, Fuliang Tang, Kunguang Wu, Zhendong Guo, Shuaishuai Huang, Yingtian Mei, Yuxing Wang, Zeyu Yang, and Shiming Gong (2024). “The Large Language Model (LLM) Bias Evaluation (Age Bias)- DIKWP Research Group International Standard Evaluation.
Caliskan, A., Bryson, J., and Narayanan, A. (2017). Semantics derived automatically from language corpora contain human-like biases.
Conneau, A., Khandelwal, K., Goyal, N., Chaudhary, V., Wenzek, G., Guzmán, F., Grave, E., Ott, M., Zettlemoyer, L., and Stoyanov, V. (2020). Unsupervised cross-lingual representation learning at scale.
Devlin, J., Chang, M.-W., Lee, K., and Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. In Burstein, J., Doran, C., and Solorio, T., editors, Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Vol. 1, pages 4171–4186, Minneapolis, Minnesota. ACL.
Jentzsch, S. and Turan, C. (2022). Gender bias in bert - measuring and analysing biases through sentiment rating in a realistic downstream classification task.
May, C., Wang, A., Bordia, S., Bowman, S. R., and Rudinger, R. (2019). On measuring social biases in sentence encoders.
Puttick, A., Rankwiler, L., Ikae, C., and Kurpicz-Briki, M. (2024). The bias detection framework: Bias detection in word embeddings and language models for european languages.
Santana, B. S., Woloszyn, V., and Wives, L. K. (2018). Is there gender bias and stereotype in portuguese word embeddings? Souza, F., Nogueira, R., and Lotufo, R. (2020). BERTimbau: Pretrained BERT Models for Brazilian Portuguese.
Taso, F., Reis, V., and Martinez, F. (2023). Sexismo no brasil: análise de um word embedding por meio de testes baseados em associação implícita.
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., and Polosukhin, I. (2023). Attention is all you need.
Yarrabelly, N., Damodaran, V., and Su, F.-G. (2024). Mitigating gender bias in contextual word embeddings.
Yucong Duan, Fuliang Tang, Kunguang Wu, Zhendong Guo, Shuaishuai Huang, Yingtian Mei, Yuxing Wang, Zeyu Yang, and Shiming Gong (2024). “The Large Language Model (LLM) Bias Evaluation (Age Bias)- DIKWP Research Group International Standard Evaluation.
Publicado
12/11/2025
Como Citar
AZEVEDO, Gabriel Kuster de; SANTANA, Brenda Salenave.
Profissões e estereótipos: avaliando o viés de gênero em versões do BERTimbau. In: ESCOLA REGIONAL DE APRENDIZADO DE MÁQUINA E INTELIGÊNCIA ARTIFICIAL DA REGIÃO SUL (ERAMIA-RS), 1. , 2025, Porto Alegre/RS.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2025
.
p. 316-319.
DOI: https://doi.org/10.5753/eramiars.2025.16678.
