Matches in SemOpenAlex for { <https://semopenalex.org/work/W3200245893> ?p ?o ?g. }
Showing items 1 to 97 of
97
with 100 items per page.
- W3200245893 abstract "Adapters are light-weight modules that allow parameter-efficient fine-tuning of pretrained models. Specialized language and task adapters have recently been proposed to facilitate cross-lingual transfer of multilingual pretrained models (Pfeiffer et al., 2020b). However, this approach requires training a separate language adapter for every language one wishes to support, which can be impractical for languages with limited data. An intuitive solution is to use a related language adapter for the new language variety, but we observe that this solution can lead to sub-optimal performance. In this paper, we aim to improve the robustness of language adapters to uncovered languages without training new adapters. We find that ensembling multiple existing language adapters makes the fine-tuned model significantly more robust to other language varieties not included in these adapters. Building upon this observation, we propose Entropy Minimized Ensemble of Adapters (EMEA), a method that optimizes the ensemble weights of the pretrained language adapters for each test sentence by minimizing the entropy of its predictions. Experiments on three diverse groups of language varieties show that our method leads to significant improvements on both named entity recognition and part-of-speech tagging across all languages." @default.
- W3200245893 created "2021-09-27" @default.
- W3200245893 creator A5037310413 @default.
- W3200245893 creator A5040523461 @default.
- W3200245893 creator A5062910836 @default.
- W3200245893 creator A5068811427 @default.
- W3200245893 date "2021-01-01" @default.
- W3200245893 modified "2023-09-27" @default.
- W3200245893 title "Efficient Test Time Adapter Ensembling for Low-resource Language Varieties" @default.
- W3200245893 cites W1987093418 @default.
- W3200245893 cites W1995875735 @default.
- W3200245893 cites W2812315852 @default.
- W3200245893 cites W2915977242 @default.
- W3200245893 cites W2963211188 @default.
- W3200245893 cites W2963341956 @default.
- W3200245893 cites W2964303773 @default.
- W3200245893 cites W2970925270 @default.
- W3200245893 cites W2971207485 @default.
- W3200245893 cites W3004076319 @default.
- W3200245893 cites W3013840636 @default.
- W3200245893 cites W3034408878 @default.
- W3200245893 cites W3035390927 @default.
- W3200245893 cites W3035579820 @default.
- W3200245893 cites W3099793224 @default.
- W3200245893 cites W3101498587 @default.
- W3200245893 cites W3105421296 @default.
- W3200245893 cites W3113057009 @default.
- W3200245893 cites W3117088788 @default.
- W3200245893 cites W3119438769 @default.
- W3200245893 cites W3119543788 @default.
- W3200245893 cites W3153597994 @default.
- W3200245893 cites W3153675281 @default.
- W3200245893 cites W3170666909 @default.
- W3200245893 cites W3173785189 @default.
- W3200245893 doi "https://doi.org/10.18653/v1/2021.findings-emnlp.63" @default.
- W3200245893 hasPublicationYear "2021" @default.
- W3200245893 type Work @default.
- W3200245893 sameAs 3200245893 @default.
- W3200245893 citedByCount "2" @default.
- W3200245893 countsByYear W32002458932022 @default.
- W3200245893 countsByYear W32002458932023 @default.
- W3200245893 crossrefType "proceedings-article" @default.
- W3200245893 hasAuthorship W3200245893A5037310413 @default.
- W3200245893 hasAuthorship W3200245893A5040523461 @default.
- W3200245893 hasAuthorship W3200245893A5062910836 @default.
- W3200245893 hasAuthorship W3200245893A5068811427 @default.
- W3200245893 hasBestOaLocation W32002458931 @default.
- W3200245893 hasConcept C104317684 @default.
- W3200245893 hasConcept C111919701 @default.
- W3200245893 hasConcept C137293760 @default.
- W3200245893 hasConcept C138885662 @default.
- W3200245893 hasConcept C154945302 @default.
- W3200245893 hasConcept C177284502 @default.
- W3200245893 hasConcept C185592680 @default.
- W3200245893 hasConcept C204321447 @default.
- W3200245893 hasConcept C2777530160 @default.
- W3200245893 hasConcept C28490314 @default.
- W3200245893 hasConcept C41008148 @default.
- W3200245893 hasConcept C41895202 @default.
- W3200245893 hasConcept C55493867 @default.
- W3200245893 hasConcept C63479239 @default.
- W3200245893 hasConcept C94922259 @default.
- W3200245893 hasConcept C9679016 @default.
- W3200245893 hasConceptScore W3200245893C104317684 @default.
- W3200245893 hasConceptScore W3200245893C111919701 @default.
- W3200245893 hasConceptScore W3200245893C137293760 @default.
- W3200245893 hasConceptScore W3200245893C138885662 @default.
- W3200245893 hasConceptScore W3200245893C154945302 @default.
- W3200245893 hasConceptScore W3200245893C177284502 @default.
- W3200245893 hasConceptScore W3200245893C185592680 @default.
- W3200245893 hasConceptScore W3200245893C204321447 @default.
- W3200245893 hasConceptScore W3200245893C2777530160 @default.
- W3200245893 hasConceptScore W3200245893C28490314 @default.
- W3200245893 hasConceptScore W3200245893C41008148 @default.
- W3200245893 hasConceptScore W3200245893C41895202 @default.
- W3200245893 hasConceptScore W3200245893C55493867 @default.
- W3200245893 hasConceptScore W3200245893C63479239 @default.
- W3200245893 hasConceptScore W3200245893C94922259 @default.
- W3200245893 hasConceptScore W3200245893C9679016 @default.
- W3200245893 hasLocation W32002458931 @default.
- W3200245893 hasLocation W32002458932 @default.
- W3200245893 hasOpenAccess W3200245893 @default.
- W3200245893 hasPrimaryLocation W32002458931 @default.
- W3200245893 hasRelatedWork W1517743118 @default.
- W3200245893 hasRelatedWork W1567338489 @default.
- W3200245893 hasRelatedWork W159132833 @default.
- W3200245893 hasRelatedWork W1978971213 @default.
- W3200245893 hasRelatedWork W2167717037 @default.
- W3200245893 hasRelatedWork W2952992889 @default.
- W3200245893 hasRelatedWork W2962966012 @default.
- W3200245893 hasRelatedWork W3107474891 @default.
- W3200245893 hasRelatedWork W3213961506 @default.
- W3200245893 hasRelatedWork W38394648 @default.
- W3200245893 isParatext "false" @default.
- W3200245893 isRetracted "false" @default.
- W3200245893 magId "3200245893" @default.
- W3200245893 workType "article" @default.