Debiasing Sentence Embedders through Contrastive Word Pairs
This addresses bias removal in sentence embeddings for NLP applications, offering a more effective solution than existing incremental methods.
The paper tackles the problem of bias in sentence embeddings, which persists when debiasing methods designed for word embeddings are applied due to nonlinearities, and proposes a new approach that removes both linear and nonlinear bias without harming downstream performance, as validated through comparisons on bias metrics.
Over the last years, various sentence embedders have been an integral part in the success of current machine learning approaches to Natural Language Processing (NLP). Unfortunately, multiple sources have shown that the bias, inherent in the datasets upon which these embedding methods are trained, is learned by them. A variety of different approaches to remove biases in embeddings exists in the literature. Most of these approaches are applicable to word embeddings and in fewer cases to sentence embeddings. It is problematic that most debiasing approaches are directly transferred from word embeddings, therefore these approaches fail to take into account the nonlinear nature of sentence embedders and the embeddings they produce. It has been shown in literature that bias information is still present if sentence embeddings are debiased using such methods. In this contribution, we explore an approach to remove linear and nonlinear bias information for NLP solutions, without impacting downstream performance. We compare our approach to common debiasing methods on classical bias metrics and on bias metrics which take nonlinear information into account.