A General Framework for Implicit and Explicit Debiasing of Distributional Word Vector Spaces
This work addresses the issue of human biases encoded in word embeddings for NLP researchers and practitioners, offering a more rigorous and adaptable solution, though it is incremental in improving existing debiasing approaches.
The authors tackled the problem of inconsistent and limited debiasing methods for distributional word vectors by introducing a general framework with explicit and implicit bias specifications, proposing three debiasing models, and developing a comprehensive evaluation framework. The result was that these models often completely removed bias without degrading semantic information across three embedding methods and enabled cross-lingual bias removal.
Distributional word vectors have recently been shown to encode many of the human biases, most notably gender and racial biases, and models for attenuating such biases have consequently been proposed. However, existing models and studies (1) operate on under-specified and mutually differing bias definitions, (2) are tailored for a particular bias (e.g., gender bias) and (3) have been evaluated inconsistently and non-rigorously. In this work, we introduce a general framework for debiasing word embeddings. We operationalize the definition of a bias by discerning two types of bias specification: explicit and implicit. We then propose three debiasing models that operate on explicit or implicit bias specifications and that can be composed towards more robust debiasing. Finally, we devise a full-fledged evaluation framework in which we couple existing bias metrics with newly proposed ones. Experimental findings across three embedding methods suggest that the proposed debiasing models are robust and widely applicable: they often completely remove the bias both implicitly and explicitly without degradation of semantic information encoded in any of the input distributional spaces. Moreover, we successfully transfer debiasing models, by means of cross-lingual embedding spaces, and remove or attenuate biases in distributional word vector spaces of languages that lack readily available bias specifications.