Counterfactual Data Augmentation for Mitigating Gender Stereotypes in Languages with Rich Morphology

By Ran Zmigrod, Sabrina J. Mielke, Hanna M. Wallach, Ryan Cotterell. 2019

344 citations recorded by the source corpus.

Gender stereotypes are manifest in most of the world’s languages and are consequently propagated or amplified by NLP systems. Although research has focused on mitigating gender stereotypes in English, the approaches that are commonly employed produce ungrammatical sentences in morphologically rich languages. We present a novel approach for converting between masculine-inflected and feminine-inflected sentences in such languages. For Spanish and Hebrew, our approach achieves F1 scores of 82% and 73% at the level of tags and accuracies of 90% and 87% at the level of forms. By evaluating our approach using four different languages, we show that, on average, it reduces gender stereotyping by a factor of 2.5 without any sacrifice to grammaticality.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Language (Technology) is Power: A Critical Survey of “Bias” in NLP (2020, neutral): “…et al. (2019); Mayfield et al. (2019); Pujari et al. (2019); Romanov et al. (2019); Ruane et al. (2019); Sedoc and Ungar (2019); Sun et al. (2019); Zmigrod et al. (2019); Hutchinson et al. (2020); Papakyriakopoulos et al. (2020); Ravfogel et al. (2020); Strengers et al. (2020); Sweeney and…”
Societal Biases in Language Generation: Progress and Challenges (2021, methodology): “Other techniques include using de-biased word embeddings (Escud´e Font and Costa-juss`a, 2019), identifying and editing out subjective words (Pryzant et al., 2020), and using Markov random ﬁelds to preserve morpho-syntactic agreement during reinﬂection (Zmigrod et al., 2019).”
Re-imagining Algorithmic Fairness in India and Beyond (2021, background): “perturbation testing [156], data augmentation [220], adversarial testing [117], and adherence to terminology guidelines for oppressed groups, such as SIGACCESS.”
Toward Gender-Inclusive Coreference Resolution (2019, methodology): “We then generate counterfactual “augmentations” of this dataset by ablating the various notions of linguistic gender described in §2.2, similar to [Zmigrod et al. 2019].”
Queens Are Powerful Too: Mitigating Gender Bias in Dialogue Generation (2019, methodology): “A straightforward solution for gender bias in embeddings is Counterfactual Data Augmentation (CDA) (Hall Maudslay et al., 2019; Zmigrod et al., 2019; Liu et al., 2019).”
Learning the Difference that Makes a Difference with Counterfactually-Augmented Data (2019, background): “Building on their work, Zmigrod et al. (2019) describe a data augmentation approach for mitigating gender stereotypes associated with animate nouns for morphologically-rich languages like Spanish and Hebrew.”
Towards Controllable Biases in Language Generation (2020, neutral): “, 2019), and debiasing across languages (Zhou et al., 2019; Zmigrod et al., 2019)).”
Predictive Biases in Natural Language Processing Models: A Conceptual Framework and Overview (2019, background): “Recently, Zmigrod et al. (2019) showed that gender-based selection bias can be addressed by data augmentation, i.e., by adding slightly altered examples to the data.”
You reap what you sow: On the Challenges of Bias Evaluation Under Multilingual Settings (2022, background): “The same 665 does not hold true for genderless languages, despite 666 the existence of the bias (Zmigrod et al., 2019).”
Measuring and Reducing Gendered Correlations in Pre-trained Models (2020, positive): “…strategies have become popular recently, in particular re-balancing (Dixon et al., 2018) and counterfactual data augmentation (CDA), which augments training data using controlled perturbations to names or demographic attributes (Hall Maudslay et al., 2019; Zhao et al., 2019; Zmigrod et al., 2019).”
Automatic Gender Identification and Reinflection in Arabic (2019, methodology): “Building on CDA, (Zmigrod et al., 2019) presented a generative model that allows conversion between masculine inﬂected and feminine inﬂected sentences in four morphologically rich languages (Hebrew, Spanish, French and Italian) with a focus on animate nouns.”
Reducing Gender Bias in Neural Machine Translation as a Domain Adaptation Problem (2020, neutral): “Recent approaches to the bias problem in NLP have involved training from scratch on artiﬁcially gender-balanced versions of the original dataset (Zhao et al., 2018; Zmigrod et al., 2019) or with de-biased embeddings (Escud´e Font and Costa-juss`a, 2019; Bolukbasi et al., 2016).”
What do Bias Measures Measure? (2021, background): “Individual re-writing tasks have curated datasets for evaluating biases in speciﬁc tasks (e.g., Habash et al. (2019) build a corpus for gender reinﬂection in Arabic, Zmigrod et al. (2019) annotate Hebrew and Spanish datasets for reinﬂection) and generally use accuracy (i.e., correct gender inﬂection”
Adhering, Steering, and Queering: Treatment of Gender in Natural Language Generation (2020, background): “Similarly, Zmigrod et al. [76] considers ﬂipping the grammatical gender of nouns and their surrounding words in non-English languages.”
Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates (2021, background): “…2018; Elazar and Goldberg, 2018; Kumar et al., 2019); (2) data-based approaches to balance the training data, e.g., counterfactual data augmentation (Zmigrod et al., 2019; Kaushik et al., 2020); (3) optimization approaches to account for worst-case scenarios, e.g., distributionally robust…”
Bias and Fairness in Large Language Models: A Survey (2023, background): “Counterfactual data augmentation (CDA) is one of the primary of these augmentation techniques (Lu et al., 2020; Qian et al., 2022; Webster et al., 2020; Zmigrod et al., 2019), replacing protected attribute words, such as gendered pronouns, to achieve a balanced dataset.”