Title: Mitigating Gender Bias in English–Spanish Machine Translation: A Parameter-Efficient Fine-Tuning Approach

Abstract: This study investigates gender bias in neural machine translation from English to Spanish and proposes a practical mitigation strategy. We first measure baseline translations and find that occupation-related sentences often skew toward stereotypical gender forms. To address this, we compile a small, balanced corpus of masculine and feminine contexts and fine-tune a high-quality baseline model using parameter-efficient adapters. The adapted model markedly improves gender agreement while maintaining overall translation quality, as measured by standard metrics. Our experiments demonstrate that gender accuracy improves substantially for feminine cases (from 43% to 69% in feminine examples), while the bias gap decreases from 0.41 to 0.06. This lightweight approach combines targeted test sentences with transparent scoring and error analysis, offering an accessible framework for detecting and mitigating gender bias in translation systems. The method is easily transferable to other language pairs that require grammatical gender agreement and is well-suited for research or production environments with limited data and computing resources.

Keywords: Gender bias, machine translation, gender agreement, bias mitigation

• E. D. Dursun, M. A. Toçoğlu, and E. Şatır. Forecasting performance of quantitative strategies with OpenAI GPT-4. Journal of Intelligent Systems with Applications, vol. 7, no. 2, pp. 24–30, 2024.
• E. Hu, Y. Shen, P. Wallis et al. LoRA: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021.
• K. Papineni, S. Roukos, T. Ward, and W.-J. Zhu. BLEU: A method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics (ACL), 2022.
• G. Stanovsky, A. Wang, C. Lin et al. Evaluating gender bias in machine translation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics (ACL 2019), 2019.
• A. Fan, S. Bhosale, H. Schwenk et al. Beyond English-centric multilingual machine translation. Journal of Machine Learning Research, vol. 22, no. 107, pp. 1–48, 2021.
• J. Pfeiffer, F. Guzman, M. Ali et al. AdapterFusion: Non-destructive task composition for transfer learning. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), 2020.
• J. Tiedemann, M. Aulamo, D. Bakshandaeva et al. Democratizing neural machine translation with OPUS-MT. Language Resources and Evaluation, vol. 58, pp. 713–755, 2023.
• J. Zhao, T. Wang, M. Yatskar et al. Men also like shopping: Reducing gender bias amplification using corpus-level constraints. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2017.
• J. Zhao, T. Wang, M. Yatskar et al. Gender bias in coreference resolution: Evaluation and debiasing methods. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT 2018), pp. 15–20, 2018.
• M. O. R. Prates, P. H. Avelar, and L. C. Lamb. Assessing gender bias in machine translation: A case study with Google Translate. Neural Computing and Applications, vol. 32, pp. 6363-6381, 2020.
• T. Sun, X. Qiu, and X. Huang. A post-editing approach to gender bias mitigation in machine translation. In Proceedings of EMNLP, 2020.
• M. Honnibal, and I. Montani, I. spaCy 2: Natural language understanding with Bloom embeddings, convolutional neural networks, and incremental parsing. Sentometrics Research, Publication ID 72, 2017.
• L. Padro, and E. Stanilovsky. FreeLing 3.0: Towards wider multilinguality. In Proceedings of the Language Resources and Evaluation Conference (LREC 2012), pp. 2473–2479, 2012.
• Y. Atayolu, and Y. Kutlu. Effect of text preprocessing methods on the performance of social media posts classification. Journal of Intelligent Systems with Applications, vol. 7, no. 1, pp. 1–6, 2024.
• M. Popovic. chrF: Character n-gram F-score for automatic machine translation evaluation. In Proceedings of the Tenth Workshop on Statistical Machine Translation (WMT 2015), pp. 392–395, Lisbon, Portugal, 2015.
• S. Mangrulkar, S. Gugger, L. Debut et al. . PEFT: State-of-the-art parameter-efficient fine-tuning methods. 2022. Retrieved from https://github.com/huggingface/peft