Attenuating Bias in Word Vectors

By Sunipa Dev, Jeff M. Phillips. 2019

161 citations recorded by the source corpus.

Word vector representations are well developed tools for various NLP and Machine Learning tasks and are known to retain significant semantic and syntactic structure of languages. But they are prone to carrying and amplifying bias which can perpetrate discrimination in various applications. In this work, we explore new simple ways to detect the most stereotypically gendered words in an embedding and remove the bias from them. We verify how names are masked carriers of gender bias and then use that as a tool to attenuate bias in embeddings. Further, we extend this property of names to show how names can be used to detect other types of bias in the embeddings such as bias based on race, ethnicity, and age.

Subfields: Artificial Intelligence, General Health Professions, Representation & Self-Supervised Learning

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 General Health Professions

In Representation & Self-Supervised Learning

Citations in the corpus

Language (Technology) is Power: A Critical Survey of “Bias” in NLP (2020, background): “…and Bhallamudi (2019); Bordia and Bowman (2019); Brunet et al. (2019); Cao and Daumé (2019); Chaloner and Maldonado (2019); Cho et al. (2019); Dev and Phillips (2019); Font and Costa-jussà (2019); Gonen and Goldberg (2019); James-Sorenson and Alvarez-Melis (2019); Kaneko and Bollegala…”
Harms of Gender Exclusivity and Challenges in Non-Binary Representation in Language Technologies (2021, neutral): “Subspace analyses Capturing a gender subspace has been useful in techniques of bias analysis and techniques in subsequent debiasing in binary gender (Dev and Phillips, 2019), especially in context-free or static representations like GloVe or word2vec.”
On Measuring and Mitigating Biased Inferences of Word Embeddings (2019, methodology): “We focus on the simple projection operator (Dev and Phillips 2019) which simply identiﬁes a subspace associated with a concept hypothesized to carry bias, and then removes that subspace from all word representations.”
Examining Gender Bias in Languages with Grammatical Gender (2019, background): “Therefore, extensive effort has been put toward analyzing and mitigating gender bias in word embeddings (Bolukbasi et al., 2016; Zhao et al., 2018b; Dev and Phillips, 2019; Ethayarajh et al., 2019).”
JW300: A Wide-Coverage Parallel Corpus for Low-Resource Languages (2019, background): “…areas, we argue that a wide-coverage collection of parallel data with known bias may in fact be valuable for research on bias in NLP (Bolukbasi et al., 2016; Caliskan et al., 2017; Dev and Phillips, 2019; Go-nen and Goldberg, 2019), especially in multilingual settings (Lauscher and Glavaˇs, 2019).”
RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models (2021, background): “…Gonen and Goldberg, 2019; Ethayarajh et al., 2019), more comprehensive evaluation frameworks (Lauscher et al., 2020a), new debiasing approaches (Dev and Phillips, 2019; Karve et al., 2019) and task-speciﬁc bias measures and resources for tasks like coreference resolution (Zhao et al., 2018),…”
deb2viz: Debiasing gender in word embedding data using subspace visualization (2020, methodology): “We also return the average scores (EQT-Avg) as computed by [8].”
Neutralizing Gender Bias in Word Embedding with Latent Disentanglement and Counterfactual Generation (2020, methodology): “ATT-Debias (Dev and Phillips, 2019) defines gender subspace with common names and proposes the subtraction and the linear projection methods based on gender subspace.”
A General Framework for Implicit and Explicit Debiasing of Distributional Word Vector Spaces (2019, methodology): “Based on the observation of (Bolukbasi et al. 2016) that in a biased vector space programmer − homemaker ≈ man − woman , Dev and Phillips (2019) proposed an analogy-based bias test: Em-bedding Quality Test (EQT).”
Bias in word embeddings (2020, background): “[27] show that names in word embeddings function as proxies of bias against social groups.”
The Geometry of Distributed Representations for Better Alignment, Attenuated Bias, and Improved Interpretability (2020, methodology): “As seen in Chapter 4, to discover biases, several lines of existing work [12, 17, 28, 105]”
Evaluating Gender Bias in Hindi-English Machine Translation (2021, methodology): “The Embedding Coherence Test (Dev and Phillips, 2019) compares the vectors of the two target sets T 1 and T 2 , averaged over all their terms, with vectors from an attribute set A .”
Socially Aware Bias Measurements for Hindi Language Representations (2021, unknown): “Both Bolukbasi et al. (2016) and Dev and Phillips (2019) propose using linear projection to debias word embeddings, but the former in addition also equalizes word pairs about the attribute (e.g., gender) axis.”
A Survey on Gender Bias in Natural Language Processing (2021, background): “Further, Dev et al. [2020] employ an orthogonal projection to gender direction [Dev and Phillips 2019] to debias contextualised embeddings and test it on a NLI task with gender-biased hypothesis pairs.”
Quantifying Societal Bias Amplification in Image Captioning (2022, methodology): “We do not initialize the LSTM with pre-computed word embeddings, as they contain bias [6, 14].”
Joint Multiclass Debiasing of Word Embeddings (2020, methodology): “on-binary class environment, which is formulated via PCA of mean shifted k-tuples (k=number of subclasses) of the denitional words. There are also other recently proposed gender bias post-processing [4,5] and pre-processing techniques [16]. In terms of existing joint multiclass debiasing techniques,”
On the Relationships Between the Grammatical Genders of Inanimate Nouns and Their Co-Occurring Adjectives and Verbs (2020, background): “Finally, many scholars have proposed a variety of computational techniques for mitigating gender norms and stereotypes in a wide range of languagebased applications (Dev and Phillips, 2019; Dinan et al., 2019; Ethayarajh et al., 2019; Hall Maudslay et al., 2019; Stanovsky et al., 2019; Tan and Celis”
On Measures of Biases and Harms in NLP (2021, background): “, 2020), and mitigate (Dev and Phillips, 2019; Ravfogel et al., 2020; Sun et al., 2019; Dev et al., 2021a) the social biases encoded by NLP models.”
AraWEAT: Multidimensional Analysis of Biases in Arabic Word Embeddings (2020, methodology): “, 2017), (2) Embedding Coherence Test (ECT) (Dev and Phillips, 2019), (3) Bias Analogy Test (BAT) (Lauscher et al.”