Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386804534> ?p ?o ?g. }
- W4386804534 endingPage "294" @default.
- W4386804534 startingPage "276" @default.
- W4386804534 abstract "Unjustified social stereotypes have lately been found to taint the predictions of NLP models. Thus, an increasing amount of research focuses on developing methods to mitigate social bias. Most proposed approaches update the parameters of models post-hoc, running the risk of forgetting the predictive task of interest. In this work, we propose a novel way of debiasing NLP models by debiasing and curating their training data. To do so, we propose an unsupervised pipeline to identify which instances in the training data mention stereotypes that tally with the stereotypes encoded in NLP models. Then we either remove or augment these problematic instances, and train NLP models on less biased data. In this pipeline, we propose three methods to excavate stereotypes encoded in models using likelihoods, attention weights and vector representations. Experiments on the tasks of natural language inference, sentiment analysis and question answering suggest that our methods are better at debiasing downstream models than existing techniques." @default.
- W4386804534 created "2023-09-17" @default.
- W4386804534 creator A5012083595 @default.
- W4386804534 creator A5027710558 @default.
- W4386804534 creator A5067470399 @default.
- W4386804534 creator A5086585292 @default.
- W4386804534 date "2023-01-01" @default.
- W4386804534 modified "2023-09-27" @default.
- W4386804534 title "Targeting the Source: Selective Data Curation for Debiasing NLP Models" @default.
- W4386804534 cites W1682403713 @default.
- W4386804534 cites W2401379394 @default.
- W4386804534 cites W2560647685 @default.
- W4386804534 cites W2795975316 @default.
- W4386804534 cites W2888161220 @default.
- W4386804534 cites W2893425640 @default.
- W4386804534 cites W2949969209 @default.
- W4386804534 cites W2950866572 @default.
- W4386804534 cites W2950888501 @default.
- W4386804534 cites W2954275542 @default.
- W4386804534 cites W2962990575 @default.
- W4386804534 cites W2963078909 @default.
- W4386804534 cites W2963524349 @default.
- W4386804534 cites W2963526187 @default.
- W4386804534 cites W2963748441 @default.
- W4386804534 cites W2963879260 @default.
- W4386804534 cites W2972413484 @default.
- W4386804534 cites W2989344603 @default.
- W4386804534 cites W2997588435 @default.
- W4386804534 cites W3034937117 @default.
- W4386804534 cites W3035241006 @default.
- W4386804534 cites W3035591180 @default.
- W4386804534 cites W3037831233 @default.
- W4386804534 cites W3046292371 @default.
- W4386804534 cites W3099635335 @default.
- W4386804534 cites W3099793224 @default.
- W4386804534 cites W3100881017 @default.
- W4386804534 cites W3103639864 @default.
- W4386804534 cites W3105042180 @default.
- W4386804534 cites W3105882417 @default.
- W4386804534 cites W3134354193 @default.
- W4386804534 cites W3155655882 @default.
- W4386804534 cites W3156204678 @default.
- W4386804534 cites W3171676064 @default.
- W4386804534 cites W3172415559 @default.
- W4386804534 cites W3176477796 @default.
- W4386804534 cites W3177189402 @default.
- W4386804534 cites W4206292552 @default.
- W4386804534 cites W4229008248 @default.
- W4386804534 cites W4241439391 @default.
- W4386804534 cites W4285152678 @default.
- W4386804534 cites W4318046968 @default.
- W4386804534 cites W4385573981 @default.
- W4386804534 doi "https://doi.org/10.1007/978-3-031-43415-0_17" @default.
- W4386804534 hasPublicationYear "2023" @default.
- W4386804534 type Work @default.
- W4386804534 citedByCount "0" @default.
- W4386804534 crossrefType "book-chapter" @default.
- W4386804534 hasAuthorship W4386804534A5012083595 @default.
- W4386804534 hasAuthorship W4386804534A5027710558 @default.
- W4386804534 hasAuthorship W4386804534A5067470399 @default.
- W4386804534 hasAuthorship W4386804534A5086585292 @default.
- W4386804534 hasConcept C119857082 @default.
- W4386804534 hasConcept C154945302 @default.
- W4386804534 hasConcept C15744967 @default.
- W4386804534 hasConcept C162324750 @default.
- W4386804534 hasConcept C187736073 @default.
- W4386804534 hasConcept C188147891 @default.
- W4386804534 hasConcept C199360897 @default.
- W4386804534 hasConcept C204321447 @default.
- W4386804534 hasConcept C2776214188 @default.
- W4386804534 hasConcept C2779458634 @default.
- W4386804534 hasConcept C2780451532 @default.
- W4386804534 hasConcept C41008148 @default.
- W4386804534 hasConcept C43521106 @default.
- W4386804534 hasConceptScore W4386804534C119857082 @default.
- W4386804534 hasConceptScore W4386804534C154945302 @default.
- W4386804534 hasConceptScore W4386804534C15744967 @default.
- W4386804534 hasConceptScore W4386804534C162324750 @default.
- W4386804534 hasConceptScore W4386804534C187736073 @default.
- W4386804534 hasConceptScore W4386804534C188147891 @default.
- W4386804534 hasConceptScore W4386804534C199360897 @default.
- W4386804534 hasConceptScore W4386804534C204321447 @default.
- W4386804534 hasConceptScore W4386804534C2776214188 @default.
- W4386804534 hasConceptScore W4386804534C2779458634 @default.
- W4386804534 hasConceptScore W4386804534C2780451532 @default.
- W4386804534 hasConceptScore W4386804534C41008148 @default.
- W4386804534 hasConceptScore W4386804534C43521106 @default.
- W4386804534 hasLocation W43868045341 @default.
- W4386804534 hasOpenAccess W4386804534 @default.
- W4386804534 hasPrimaryLocation W43868045341 @default.
- W4386804534 hasRelatedWork W2015950787 @default.
- W4386804534 hasRelatedWork W2020540721 @default.
- W4386804534 hasRelatedWork W2578916128 @default.
- W4386804534 hasRelatedWork W2961085424 @default.
- W4386804534 hasRelatedWork W2992516105 @default.
- W4386804534 hasRelatedWork W4306674287 @default.
- W4386804534 hasRelatedWork W4306704608 @default.
- W4386804534 hasRelatedWork W4377864593 @default.