Matches in SemOpenAlex for { <https://semopenalex.org/work/W3201217479> ?p ?o ?g. }
- W3201217479 abstract "Contextual representations learned by language models can often encode undesirable attributes, like demographic associations of the users, while being trained for an unrelated target task. We aim to scrub such undesirable attributes and learn fair representations while maintaining performance on the target task. In this paper, we present an adversarial learning framework “Adversarial Scrubber” (AdS), to debias contextual representations. We perform theoretical analysis to show that our framework converges without leaking demographic information under certain conditions. We extend previous evaluation techniques by evaluating debiasing performance using Minimum Description Length (MDL) probing. Experimental evaluations on 8 datasets show that AdS generates representations with minimal information about demographic attributes while being maximally informative about the target task." @default.
- W3201217479 created "2021-09-27" @default.
- W3201217479 creator A5017183605 @default.
- W3201217479 creator A5023432391 @default.
- W3201217479 creator A5034885122 @default.
- W3201217479 creator A5041254552 @default.
- W3201217479 creator A5050408389 @default.
- W3201217479 creator A5090095198 @default.
- W3201217479 date "2021-01-01" @default.
- W3201217479 modified "2023-09-25" @default.
- W3201217479 title "Adversarial Scrubbing of Demographic Information for Text Classification" @default.
- W3201217479 cites W1893888233 @default.
- W3201217479 cites W2025403586 @default.
- W3201217479 cites W2099813784 @default.
- W3201217479 cites W2132119275 @default.
- W3201217479 cites W2162670686 @default.
- W3201217479 cites W2252024402 @default.
- W3201217479 cites W2262564268 @default.
- W3201217479 cites W2346625195 @default.
- W3201217479 cites W2371804352 @default.
- W3201217479 cites W2547875792 @default.
- W3201217479 cites W2572403455 @default.
- W3201217479 cites W2575864632 @default.
- W3201217479 cites W2739810148 @default.
- W3201217479 cites W2786672974 @default.
- W3201217479 cites W2791170418 @default.
- W3201217479 cites W2802105481 @default.
- W3201217479 cites W2888161220 @default.
- W3201217479 cites W2945985835 @default.
- W3201217479 cites W2949678053 @default.
- W3201217479 cites W2950018712 @default.
- W3201217479 cites W2952328691 @default.
- W3201217479 cites W2962990575 @default.
- W3201217479 cites W2963078909 @default.
- W3201217479 cites W2963116854 @default.
- W3201217479 cites W2963341956 @default.
- W3201217479 cites W2963879260 @default.
- W3201217479 cites W2964235839 @default.
- W3201217479 cites W2970862333 @default.
- W3201217479 cites W2971307358 @default.
- W3201217479 cites W2972568911 @default.
- W3201217479 cites W3034458735 @default.
- W3201217479 cites W3037831233 @default.
- W3201217479 cites W3100355408 @default.
- W3201217479 cites W3101004475 @default.
- W3201217479 cites W3104617516 @default.
- W3201217479 cites W3148140980 @default.
- W3201217479 cites W3157576094 @default.
- W3201217479 cites W3159900299 @default.
- W3201217479 cites W3181414820 @default.
- W3201217479 cites W9292421 @default.
- W3201217479 doi "https://doi.org/10.18653/v1/2021.emnlp-main.43" @default.
- W3201217479 hasPublicationYear "2021" @default.
- W3201217479 type Work @default.
- W3201217479 sameAs 3201217479 @default.
- W3201217479 citedByCount "1" @default.
- W3201217479 countsByYear W32012174792022 @default.
- W3201217479 crossrefType "proceedings-article" @default.
- W3201217479 hasAuthorship W3201217479A5017183605 @default.
- W3201217479 hasAuthorship W3201217479A5023432391 @default.
- W3201217479 hasAuthorship W3201217479A5034885122 @default.
- W3201217479 hasAuthorship W3201217479A5041254552 @default.
- W3201217479 hasAuthorship W3201217479A5050408389 @default.
- W3201217479 hasAuthorship W3201217479A5090095198 @default.
- W3201217479 hasBestOaLocation W32012174791 @default.
- W3201217479 hasConcept C104317684 @default.
- W3201217479 hasConcept C119857082 @default.
- W3201217479 hasConcept C137293760 @default.
- W3201217479 hasConcept C154945302 @default.
- W3201217479 hasConcept C15744967 @default.
- W3201217479 hasConcept C162324750 @default.
- W3201217479 hasConcept C175154964 @default.
- W3201217479 hasConcept C185592680 @default.
- W3201217479 hasConcept C187736073 @default.
- W3201217479 hasConcept C188147891 @default.
- W3201217479 hasConcept C204321447 @default.
- W3201217479 hasConcept C2779458634 @default.
- W3201217479 hasConcept C2780451532 @default.
- W3201217479 hasConcept C37736160 @default.
- W3201217479 hasConcept C41008148 @default.
- W3201217479 hasConcept C55493867 @default.
- W3201217479 hasConcept C66746571 @default.
- W3201217479 hasConceptScore W3201217479C104317684 @default.
- W3201217479 hasConceptScore W3201217479C119857082 @default.
- W3201217479 hasConceptScore W3201217479C137293760 @default.
- W3201217479 hasConceptScore W3201217479C154945302 @default.
- W3201217479 hasConceptScore W3201217479C15744967 @default.
- W3201217479 hasConceptScore W3201217479C162324750 @default.
- W3201217479 hasConceptScore W3201217479C175154964 @default.
- W3201217479 hasConceptScore W3201217479C185592680 @default.
- W3201217479 hasConceptScore W3201217479C187736073 @default.
- W3201217479 hasConceptScore W3201217479C188147891 @default.
- W3201217479 hasConceptScore W3201217479C204321447 @default.
- W3201217479 hasConceptScore W3201217479C2779458634 @default.
- W3201217479 hasConceptScore W3201217479C2780451532 @default.
- W3201217479 hasConceptScore W3201217479C37736160 @default.
- W3201217479 hasConceptScore W3201217479C41008148 @default.
- W3201217479 hasConceptScore W3201217479C55493867 @default.
- W3201217479 hasConceptScore W3201217479C66746571 @default.
- W3201217479 hasLocation W32012174791 @default.