📍 virtual MZES, Mannheim
📆 June 01, 2022
In this tutorial, I will show you how the R package sweater can be used to detect biases in word embeddings. The package provides highly optimized functions to calculate the following bias metrics: mean average cosine similarity, relative norm distance, SemAxis, normalized association score, relative negative sentiment bias, embedding coherence test and word embedding association test. Using two public available word embeddings trained on media content, I am going to demonstrate how sweater can be used to study implicit gender and racial biases.
📝 Slides
👤 Chung-hong Chan is a Research Fellow at the Mannheim Center for European Social Research (MZES), University of Mannheim.