Using Adversarial Debiasing to Remove Bias from Word Embeddings

Dana Kenna

Word Embeddings have been shown to contain the societal biases present in the original corpora.Existing methods to deal with this problem have been shown to only remove superficial biases. Themethod ofAdversarial Debiasingwas presumed to be similarly superficial, but this is was not verifiedin previous works. Using the experiments that demonstrated the shallow removal in other methods, Ishow results that suggestAdversarial Debiasingis more effective at removing bias and thus motivatefurther investigation on the utility ofAdversarial Debiasing.

Knowledge Graph

arrow_drop_up

Comments

Sign up or login to leave a comment