Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387079108> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4387079108 endingPage "126851" @default.
- W4387079108 startingPage "126851" @default.
- W4387079108 abstract "The bias in machine learning models has gained increasing attention in recent years, as these models can reflect and even amplify biases present in the data used to train them. One approach to mitigating bias is identifying and down-weight features that contribute disproportionately to model predictions, which can be accomplished using saliency techniques. Current debiasing methods often lead to the loss of contextual information, where the model tends to respond incorrectly even when the gender information is present in the context; hence, even though the bias reduces, performance (coreference resolution, fluency) also reduces. This paper explores data augmentation and saliency techniques to mitigate bias in natural language generation. Specifically, we explore applying the saliency technique of SHAP (SHapley Additive exPlanations) over a model trained on debiasing using data augmentation (switching gendered words with counterparts) and then applying hard debiasing to remove the influential biased token. We build a dialogue context test setup to evaluate bias and context relevance using the presence of gendered words in the model-generated responses. The response is evaluated based on the gender information from context to ensure the model follows the gender in context. We demonstrate that this approach can effectively reduce the impact of biased features on model predictions while preserving overall model accuracy. Additionally, we discuss potential limitations and future directions for research in this area. Our findings suggest that saliency offers an avenue to address machine learning bias." @default.
- W4387079108 created "2023-09-28" @default.
- W4387079108 creator A5009725385 @default.
- W4387079108 creator A5078987747 @default.
- W4387079108 creator A5085370631 @default.
- W4387079108 date "2023-09-01" @default.
- W4387079108 modified "2023-10-14" @default.
- W4387079108 title "Saliency Guided Debiasing: Detecting and mitigating biases in LMs using feature attribution" @default.
- W4387079108 cites W1787224781 @default.
- W4387079108 cites W2066255970 @default.
- W4387079108 cites W2282821441 @default.
- W4387079108 cites W2754517384 @default.
- W4387079108 cites W2789970635 @default.
- W4387079108 cites W2979463066 @default.
- W4387079108 cites W2990751682 @default.
- W4387079108 cites W3101609372 @default.
- W4387079108 cites W3104128335 @default.
- W4387079108 cites W3119558324 @default.
- W4387079108 cites W3135734416 @default.
- W4387079108 cites W3153120079 @default.
- W4387079108 cites W3181414820 @default.
- W4387079108 cites W3211022409 @default.
- W4387079108 cites W4205201623 @default.
- W4387079108 cites W4206256378 @default.
- W4387079108 cites W4210736086 @default.
- W4387079108 cites W4223894231 @default.
- W4387079108 cites W4280570554 @default.
- W4387079108 cites W4287204327 @default.
- W4387079108 cites W4306888994 @default.
- W4387079108 doi "https://doi.org/10.1016/j.neucom.2023.126851" @default.
- W4387079108 hasPublicationYear "2023" @default.
- W4387079108 type Work @default.
- W4387079108 citedByCount "0" @default.
- W4387079108 crossrefType "journal-article" @default.
- W4387079108 hasAuthorship W4387079108A5009725385 @default.
- W4387079108 hasAuthorship W4387079108A5078987747 @default.
- W4387079108 hasAuthorship W4387079108A5085370631 @default.
- W4387079108 hasConcept C119857082 @default.
- W4387079108 hasConcept C138268822 @default.
- W4387079108 hasConcept C138885662 @default.
- W4387079108 hasConcept C151730666 @default.
- W4387079108 hasConcept C154945302 @default.
- W4387079108 hasConcept C15744967 @default.
- W4387079108 hasConcept C158154518 @default.
- W4387079108 hasConcept C17744445 @default.
- W4387079108 hasConcept C180747234 @default.
- W4387079108 hasConcept C199539241 @default.
- W4387079108 hasConcept C204321447 @default.
- W4387079108 hasConcept C2776401178 @default.
- W4387079108 hasConcept C2779343474 @default.
- W4387079108 hasConcept C2779458634 @default.
- W4387079108 hasConcept C28076734 @default.
- W4387079108 hasConcept C41008148 @default.
- W4387079108 hasConcept C41895202 @default.
- W4387079108 hasConcept C77805123 @default.
- W4387079108 hasConcept C86803240 @default.
- W4387079108 hasConceptScore W4387079108C119857082 @default.
- W4387079108 hasConceptScore W4387079108C138268822 @default.
- W4387079108 hasConceptScore W4387079108C138885662 @default.
- W4387079108 hasConceptScore W4387079108C151730666 @default.
- W4387079108 hasConceptScore W4387079108C154945302 @default.
- W4387079108 hasConceptScore W4387079108C15744967 @default.
- W4387079108 hasConceptScore W4387079108C158154518 @default.
- W4387079108 hasConceptScore W4387079108C17744445 @default.
- W4387079108 hasConceptScore W4387079108C180747234 @default.
- W4387079108 hasConceptScore W4387079108C199539241 @default.
- W4387079108 hasConceptScore W4387079108C204321447 @default.
- W4387079108 hasConceptScore W4387079108C2776401178 @default.
- W4387079108 hasConceptScore W4387079108C2779343474 @default.
- W4387079108 hasConceptScore W4387079108C2779458634 @default.
- W4387079108 hasConceptScore W4387079108C28076734 @default.
- W4387079108 hasConceptScore W4387079108C41008148 @default.
- W4387079108 hasConceptScore W4387079108C41895202 @default.
- W4387079108 hasConceptScore W4387079108C77805123 @default.
- W4387079108 hasConceptScore W4387079108C86803240 @default.
- W4387079108 hasFunder F4320320767 @default.
- W4387079108 hasFunder F4320334764 @default.
- W4387079108 hasLocation W43870791081 @default.
- W4387079108 hasOpenAccess W4387079108 @default.
- W4387079108 hasPrimaryLocation W43870791081 @default.
- W4387079108 hasRelatedWork W1594011529 @default.
- W4387079108 hasRelatedWork W1964738998 @default.
- W4387079108 hasRelatedWork W2155069789 @default.
- W4387079108 hasRelatedWork W2379250586 @default.
- W4387079108 hasRelatedWork W2796868841 @default.
- W4387079108 hasRelatedWork W2961085424 @default.
- W4387079108 hasRelatedWork W2963526187 @default.
- W4387079108 hasRelatedWork W3031383011 @default.
- W4387079108 hasRelatedWork W3183896112 @default.
- W4387079108 hasRelatedWork W4306674287 @default.
- W4387079108 isParatext "false" @default.
- W4387079108 isRetracted "false" @default.
- W4387079108 workType "article" @default.