Gender bias estimation and mitigation techniques in word embeddings lack an understanding of their generalization capabilities. In this work, we complement prior research by comparing in a systematic way four gender bias metrics (Word Embedding Association Test, Relative Negative Sentiment Bias, Embedding Coherence Test and Bias Analogy Test), two types of projection-based gender mitigation strategies (hard- and soft-debiasing) on three well-known word embedding representations (Word2Vec, FastText and Glove). The experiments have shown that the considered word embeddings are consistent between them but the debiasing techniques are inconsistent across the different metrics, also highlighting the potential risk of unintended bias after the mitigation strategies.
Fersini, E., Candelieri, A., Pastore, L. (2023). On the Generalization of Projection-Based Gender Debiasing inWord Embedding. In International Conference Recent Advances in Natural Language Processing, RANLP (pp.336-343). Incoma Ltd [10.26615/978-954-452-092-2_038].
On the Generalization of Projection-Based Gender Debiasing inWord Embedding
Fersini, E;Candelieri, A;
2023
Abstract
Gender bias estimation and mitigation techniques in word embeddings lack an understanding of their generalization capabilities. In this work, we complement prior research by comparing in a systematic way four gender bias metrics (Word Embedding Association Test, Relative Negative Sentiment Bias, Embedding Coherence Test and Bias Analogy Test), two types of projection-based gender mitigation strategies (hard- and soft-debiasing) on three well-known word embedding representations (Word2Vec, FastText and Glove). The experiments have shown that the considered word embeddings are consistent between them but the debiasing techniques are inconsistent across the different metrics, also highlighting the potential risk of unintended bias after the mitigation strategies.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
