Abstract
Publicly available off-the-shelf word embeddings that are often used in productive applications for natural language processing have been proven to be biased. We have previously shown that this bias can come in different forms, depending on the language and the cultural context. In this work, we extend our previous work and further investigate how bias varies in different languages. We examine Italian and Swedish word embeddings for gender and origin bias, and demonstrate how an origin bias concerning local migration groups in Switzerland is included in German word embeddings. We propose BiasWords, a method to automatically detect new forms of bias. Finally, we discuss how cultural and language aspects are relevant to the impact of bias on the application and to potential mitigation measures.
Highlights
Algorithms and data-based applications are highly sensitive to bias in the underlying training data and can contain risks of discrimination for different groups of the society
We confirmed a bias for the WEAT5 experiment for Italian and Swedish word embeddings
We investigated on bias detection on regional languages— in particular for Romansh
Summary
Algorithms and data-based applications are highly sensitive to bias in the underlying training data and can contain risks of discrimination for different groups of the society. Learning from existing real-world data and making smart decisions based on such data amplifies cultural stereotypes (Barocas and Selbst, 2016) due to historical bias encoded inside the data. It is, relevant to provide metrics to measure and mitigate the bias in training data (Sun et al, 2019)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have