AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Understanding the Origins of Bias in Word Embeddings

Marc-Etienne Brunet, Colleen Alkalay-Houlihan, Ashton Anderson, Richard Zemel

ICML 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Multilingual language representation pretraining · 106 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use influence-function estimates to select documents for removalGloVe bias attribution · Wiki and NYT, WEAT1 and WEAT2correlation r^2 with retraining results: 0.986; 0.993; 0.995; 0.997Improved
Select the 10 most bias-increasing documents using the proposed method instead of PPMIGloVe, Wiki WEAT1 · Wiki WEAT1, remove 10 documentsWEAT effect size change: -4% -> -40%Improved
Remove the 10k most bias-contributing documents identified for GloVeword2vec · NYT WEAT1WEAT effect size: 1.35 -> 0.11Hurt
Remove randomly selected document setsGloVe embeddings · Wiki and NYT, WEAT1 and WEAT2significant difference from baseline bias: For 36 random perturbation sets, only 2 differed significantly (p < 0.05).No clear effect
Remove documents targeted by the proposed methodGloVe embeddings · Wiki and NYT, WEAT1 and WEAT2significant difference from baseline bias: For 40 targeted perturbation sets, only 2 did not significantly differ from their respective baselines.Improved
Remove documents influential for GloVeGloVe TOP-1 analogy performance · Perturbed embeddingsTOP-1 analogy performanceNo clear effect
Remove GloVe-identified bias-influencing documentsword2vec TOP-1 analogy performance · NYTTOP-1 analogy performanceNo clear effect
Remove documents containing rare WEAT wordsWEAT sensitivity to corpus perturbations · NYT WEAT1WEAT effect size: r^2 = 0.828; 0.07%Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy