Answers are grounded in this paper and cite their sources.
The problem of gender bias in Natural Language Processing (NLP) models has become a growing concern in the NLP community in recent years. Due to the fact that the texts on which models are trained often contain stereotypes and prejudices, different types of NLP models, regardless of the task and learning algorithms, demonstrate social biases in terms of gender, race, and religion. Word embeddings (WE) as a very common framework in NLP were shown to reproduce various prejudices as well and gender bias, in particular. Existing research on gender bias in word embeddings often focus on English language models and there is no such research for WE for Russian. In this work, word embeddings for Russian language were analyzed in terms of gender bias for the first time. Using Word Embedding Association Test method and an extended list of analyzed word categories, it was shown that Russian language word embeddings preserve and reproduce gender bias in various topics.
No takes yet. Share an insight, caveat, or question.
Alžběta Peštová (2021) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: