Matches in SemOpenAlex for { <https://semopenalex.org/work/W3155655882> ?p ?o ?g. }
- W3155655882 abstract "In comparison to the numerous debiasing methods proposed for the static non-contextualised word embeddings, the discriminative biases in contextualised embeddings have received relatively little attention. We propose a fine-tuning method that can be applied at token- or sentence-levels to debias pre-trained contextualised embeddings. Our proposed method can be applied to any pre-trained contextualised embedding model, without requiring to retrain those models. Using gender bias as an illustrative example, we then conduct a systematic study using several state-of-the-art (SoTA) contextualised representations on multiple benchmark datasets to evaluate the level of biases encoded in different contextualised embeddings before and after debiasing using the proposed method. We find that applying token-level debiasing for all tokens and across all layers of a contextualised embedding model produces the best performance. Interestingly, we observe that there is a trade-off between creating an accurate vs. unbiased contextualised embedding model, and different contextualised embedding models respond differently to this trade-off." @default.
- W3155655882 created "2021-04-26" @default.
- W3155655882 creator A5005531754 @default.
- W3155655882 creator A5073503574 @default.
- W3155655882 date "2021-01-01" @default.
- W3155655882 modified "2023-10-14" @default.
- W3155655882 title "Debiasing Pre-trained Contextualised Embeddings" @default.
- W3155655882 cites W131533222 @default.
- W3155655882 cites W1599016936 @default.
- W3155655882 cites W2130158090 @default.
- W3155655882 cites W2250539671 @default.
- W3155655882 cites W2251939518 @default.
- W3155655882 cites W2396767181 @default.
- W3155655882 cites W2483215953 @default.
- W3155655882 cites W2888161220 @default.
- W3155655882 cites W2889624842 @default.
- W3155655882 cites W2893425640 @default.
- W3155655882 cites W2897630418 @default.
- W3155655882 cites W2913897682 @default.
- W3155655882 cites W2926555354 @default.
- W3155655882 cites W2949969209 @default.
- W3155655882 cites W2950866572 @default.
- W3155655882 cites W2963078909 @default.
- W3155655882 cites W2963310665 @default.
- W3155655882 cites W2963341956 @default.
- W3155655882 cites W2963446520 @default.
- W3155655882 cites W2963457723 @default.
- W3155655882 cites W2963526187 @default.
- W3155655882 cites W2963612262 @default.
- W3155655882 cites W2963846996 @default.
- W3155655882 cites W2963879260 @default.
- W3155655882 cites W2965373594 @default.
- W3155655882 cites W2970583189 @default.
- W3155655882 cites W2970800693 @default.
- W3155655882 cites W2971015127 @default.
- W3155655882 cites W2972413484 @default.
- W3155655882 cites W2972572477 @default.
- W3155655882 cites W2978017171 @default.
- W3155655882 cites W2996428491 @default.
- W3155655882 cites W2997588435 @default.
- W3155655882 cites W3013571468 @default.
- W3155655882 cites W3019416653 @default.
- W3155655882 cites W3035102548 @default.
- W3155655882 cites W3035241006 @default.
- W3155655882 cites W3037697022 @default.
- W3155655882 cites W3104033643 @default.
- W3155655882 cites W3105882417 @default.
- W3155655882 cites W3118009250 @default.
- W3155655882 cites W2525127255 @default.
- W3155655882 doi "https://doi.org/10.18653/v1/2021.eacl-main.107" @default.
- W3155655882 hasPublicationYear "2021" @default.
- W3155655882 type Work @default.
- W3155655882 sameAs 3155655882 @default.
- W3155655882 citedByCount "14" @default.
- W3155655882 countsByYear W31556558822021 @default.
- W3155655882 countsByYear W31556558822022 @default.
- W3155655882 countsByYear W31556558822023 @default.
- W3155655882 crossrefType "proceedings-article" @default.
- W3155655882 hasAuthorship W3155655882A5005531754 @default.
- W3155655882 hasAuthorship W3155655882A5073503574 @default.
- W3155655882 hasBestOaLocation W31556558821 @default.
- W3155655882 hasConcept C119857082 @default.
- W3155655882 hasConcept C13280743 @default.
- W3155655882 hasConcept C154945302 @default.
- W3155655882 hasConcept C15744967 @default.
- W3155655882 hasConcept C185798385 @default.
- W3155655882 hasConcept C204321447 @default.
- W3155655882 hasConcept C205649164 @default.
- W3155655882 hasConcept C2777530160 @default.
- W3155655882 hasConcept C2779458634 @default.
- W3155655882 hasConcept C38652104 @default.
- W3155655882 hasConcept C41008148 @default.
- W3155655882 hasConcept C41608201 @default.
- W3155655882 hasConcept C48145219 @default.
- W3155655882 hasConcept C77805123 @default.
- W3155655882 hasConcept C80444323 @default.
- W3155655882 hasConcept C97931131 @default.
- W3155655882 hasConceptScore W3155655882C119857082 @default.
- W3155655882 hasConceptScore W3155655882C13280743 @default.
- W3155655882 hasConceptScore W3155655882C154945302 @default.
- W3155655882 hasConceptScore W3155655882C15744967 @default.
- W3155655882 hasConceptScore W3155655882C185798385 @default.
- W3155655882 hasConceptScore W3155655882C204321447 @default.
- W3155655882 hasConceptScore W3155655882C205649164 @default.
- W3155655882 hasConceptScore W3155655882C2777530160 @default.
- W3155655882 hasConceptScore W3155655882C2779458634 @default.
- W3155655882 hasConceptScore W3155655882C38652104 @default.
- W3155655882 hasConceptScore W3155655882C41008148 @default.
- W3155655882 hasConceptScore W3155655882C41608201 @default.
- W3155655882 hasConceptScore W3155655882C48145219 @default.
- W3155655882 hasConceptScore W3155655882C77805123 @default.
- W3155655882 hasConceptScore W3155655882C80444323 @default.
- W3155655882 hasConceptScore W3155655882C97931131 @default.
- W3155655882 hasLocation W31556558821 @default.
- W3155655882 hasLocation W31556558822 @default.
- W3155655882 hasOpenAccess W3155655882 @default.
- W3155655882 hasPrimaryLocation W31556558821 @default.
- W3155655882 hasRelatedWork W159132833 @default.
- W3155655882 hasRelatedWork W2886767593 @default.
- W3155655882 hasRelatedWork W3123930738 @default.