Gender bias in word embeddings of different languages

Gender bias in word embeddings of different languages

Raijmakers, Thijs (TU Delft Electrical Engineering, Mathematics and Computer Science)

Viering, T.J. (mentor)
Makrodimitris, S. (mentor)
Naseri Jahfari, A. (mentor)
Tax, D.M.J. (mentor)
Loog, M. (mentor)

Delft University of Technology

Computer Science and Engineering

CSE3000 Research Project

2020-06-21

Word embeddings are useful for various applications, such as sentiment classification (Tang et al., 2014), word translation (Xing, Wang, Liu, & Lin, 2015) and résumé parsing (Nasser, Sreejith, & Irshad, 2018). Previous research has determined that word embeddings contain gender bias, which can be problematic in certain applications such as résumé parsing. This research has addressed the question whether gender bias is present in word embeddings of different languages. Gender bias has been measured on word embedding of 26 different lan- guages with the help of the Word Embedding Association Test by Caliskan, Bryson, and Narayanan (2017). The results show that most of the tested languages seem to have bias towards male, while a few languages seem to have a bias towards female. This result is in line with previous literature.

Natural Language Processing
gender bias
Word embedding
WEAT
language

http://resolver.tudelft.nl/uuid:ea118174-0a65-438a-a417-8461e246499a

Student theses

bachelor thesis

© 2020 Thijs Raijmakers