Estimating Grammatical Gender Directions in Contextual Embeddings under Controlled and Natural Contexts
For researchers working on gender bias in NLP, this work provides a method to isolate grammatical gender in contextual embeddings, but the results are incremental and limited to Spanish.
This paper presents the first attempt to disentangle grammatical gender from semantic contamination in contextual embeddings for Spanish, using controlled templates and natural contexts. The centroid estimator achieves better performance than SVM and LDA in suppressing grammatical gender leakage while preserving semantic gender distinctions.
Contextual language models conflate grammatical gender and social semantic bias in gendered languages such as Spanish. Existing gender debiasing approaches only operate on static word embeddings leaving contextual representations unexplored for this two dimensional gender disentanglement. To address the this issue, we make the first attempt to disentangle grammatical gender from semantic contamination for contextual embeddings. We construct both controlled templates and natural Wikipedia contexts to build balanced datasets of inanimate nouns, and design a framework equipped with centroid, Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) gender direction estimators as well as contamination-aware weighting strategies. A set of dual-objective evaluation metrics is proposed to balance the suppression of grammatical gender leakage on inanimate nouns and the preservation of semantic gender distinctions for occupation terms. The results reveal that unweighted controlled contexts yield the purest grammatical gender direction, and the centroid estimator achieves better performance than discriminative baselines.