A systematic pipeline for diagnosing and reducing gender-stereotype bias in Japanese PLMs for sentiment analysis
Frontiers in Artificial Intelligence · 03.06.2026 · 9
DOI: 10.3389/frai.2026.1851069 ↗Kurz gefasst
Fachartikel in der Zeitschrift Frontiers in Artificial Intelligence, online erschienen am 03.06.2026. Verfasst von Yang Liu und 1 weiteren Autor:innen. Beteiligte Einrichtung: College of Eastern Language and Culture, Harbin Normal University, Heilongjiang, Harbin, China. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Pre-trained language models (PLMs) are widely used in sentiment analysis, but they may inherit gender-stereotypical bias from large-scale text corpora and transfer such bias to downstream sentiment predictions. Despite growing attention to gender-stereotypical bias in PLMs, existing studies predominantly focus on English corpora and static word embeddings, limiting understanding of how such bias affects sentiment analysis models and the effectiveness of mitigation strategies. In this study, we present a three-stage task-oriented pipeline for diagnosing, mitigating, and evaluating gender-stereotypical bias in Japanese PLMs for sentiment analysis tasks. Specifically, the proposed framework diagnoses bias based on pro-stereotypical (PS), anti-stereotypical (AS), and neutral test sets, which are constructed to compare stereotype-aligned, stereotype-violating, and gender-neutral contexts under the same sentiment analysis setting. We further introduce two complementary evaluation measures, the Stereotype Bias Index (SBI) and Gender Sentiment Bias (GSB), to quantify stereotype-level bias between PS and AS samples, as well as sentiment prediction differences among male, female, and neutral groups. To mitigate bias, the framework performs debiasing fine-tuning using gender-swapped training data and then quantitatively evaluates bias reduction while monitoring sentiment classification performance. Experimental results on three Japanese BERT-based sentiment analysis models demonstrate that the proposed pipeline substantially reduces gender-stereotypical bias. For the SBI metric, the bias magnitude is reduced by 97.4%, 70.0%, and 76.9% for Tohoku BERT BASE , Tohoku BERT BASE (chABSA), and Tohoku BERT BASE (JSPD), respectively. For the GSB metric, the bias magnitude is consistently reduced across gender-group comparisons, with reduction rates ranging from 33.3% to 98.8%. Meanwhile, sentiment classification performance is maintained or slightly improved after debiasing.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- College of Eastern Language and Culture, Harbin Normal University, Heilongjiang, Harbin, China.
- College of Art, Northeast Agricultural University, Heilongjiang, Harbin, China.
Zitieren
APA
Liu, Y., & Li, Z. (2026). A systematic pipeline for diagnosing and reducing gender-stereotype bias in Japanese PLMs for sentiment analysis. Frontiers in Artificial Intelligence, 9. https://doi.org/10.3389/frai.2026.1851069
Vancouver
Liu Y, Li Z. A systematic pipeline for diagnosing and reducing gender-stereotype bias in Japanese PLMs for sentiment analysis. Frontiers in Artificial Intelligence. 2026;9. doi:10.3389/frai.2026.1851069
BibTeX
@article{Liu2026,
author = {Liu, Yang and Li, Ziyang},
title = {A systematic pipeline for diagnosing and reducing gender-stereotype bias in Japanese PLMs for sentiment analysis},
journal = {Frontiers in Artificial Intelligence},
year = {2026},
volume = {9},
doi = {10.3389/frai.2026.1851069}
}
EUVIMED