Are We Consistently Biased? Multidimensional Analysis of Biases in Distributional Word Vectors

Proceedings of the Eighth Joint Conference on Lexical and Computational Semantics (*SEM), pages 85–91 Minneapolis, June 6–7, 2019. c⃝2019 Association for Computational Linguistics 85 Are We Consistently Biased? Multidimensional Analysis of Biases in Distributional Word Vectors Anne Lauscher and Goran Glavaˇs Data and Web Science Research Group University of Mannheim Mannheim, Germany {anne, goran}@informatik.uni-mannheim.de Abstract Word embeddings have recently been shown to reflect many of the pronounced societal bi- ases (e.g., gender bias or racial bias). Existing studies are, however, limited in scope and do not investigate the consistency of biases across relevant dimensions like embedding models, types of texts, and different languages. In this work, we present a systematic study of biases encoded in distributional word vector spaces: we analyze how consistent the bias effects are across languages, corpora, and embedding models. Furthermore, we analyze the cross- lingual biases encoded in bilingual embedding spaces, indicative of the effects of bias transfer encompassed in cross-lingual transfer of NLP models. Our study yields some unexpected findings, e.g., that biases can be emphasized or downplayed by different embedding models or that user-generated content may be less biased than encyclopedic text. We hope our work cat- alyzes bias research in NLP and informs the development of bias reduction techniques. 1 Introduction Recent work demonstrated that word embeddings induced from large text collections encode many human biases (e.g., Bolukbasi et al., 2016; Caliskan et al., 2017). This finding is not particularly surpris- ing given that (1) we are likely project our biases in the text that we produce and (2) these biases in text are bound to be encoded in word vectors due to the distributional nature (Harris, 1954) of the word em- bedding models (Mikolov et al., 2013a; Pennington et al., 2014; Bojanowski et al., 2017). For illustra- tion, consider the famous analogy-based gender bias example from Bolukbasi et al. (2016): “Man is to computer programmer as woman is to home- maker”. This bias will be reflected in the text (i.e., the word man will co-occur more often with words like programmer or engineer, whereas woman will more often appear next to homemaker or nurse), and will, in turn, be captured by word embeddings built from such biased texts. While biases encoded in word embeddings can be a useful data source for diachronic analyses of societal biases (e.g., Garg et al., 2018), they may cause ethical problems for many downstream applications and NLP models. In order to measure the extent to which various societal biases are captured by word embeddings, Caliskan et al. (2017) proposed the Word Embed- ding Association Test (WEAT). WEAT measures semantic similarity, computed through word em- beddings, between two sets of target words (e.g., insects vs. flowers) and two sets of attribute words (e.g., pleasant vs. unpleasant words). While they test a number of biases, the analysis is limited in scope to English as the only language, GloVe (Pen- nington et al., 2014) as the embedding model, and Common Crawl as the type of text. Following the same methodology, …

Content truncated to 500 words