FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3106156541> ?p ?o ?g. }

• W3106156541 abstract "Reading comprehension models often overfit to nuances of training datasets and fail at adversarial evaluation. Training with adversarially augmented dataset improves robustness against those adversarial attacks but hurts generalization of the models. In this work, we present several effective adversaries and automated data augmentation policy search methods with the goal of making reading comprehension models more robust to adversarial evaluation, but also improving generalization to the source domain as well as new domains and languages. We first propose three new methods for generating QA adversaries, that introduce multiple points of confusion within the context, show dependence on insertion location of the distractor, and reveal the compounding effect of mixing adversarial strategies with syntactic and semantic paraphrasing methods. Next, we find that augmenting the training datasets with uniformly sampled adversaries improves robustness to the adversarial attacks but leads to decline in performance on the original unaugmented dataset. We address this issue via RL and more efficient Bayesian policy search methods for automatically learning the best augmentation policy combinations of the transformation probability for each adversary in a large search space. Using these learned policies, we show that adversarial training can lead to significant improvements in in-domain, out-of-domain, and cross-lingual (German, Russian, Turkish) generalization." @default.
• W3106156541 created "2020-11-23" @default.
• W3106156541 creator A5001987532 @default.
• W3106156541 creator A5048025332 @default.
• W3106156541 date "2020-01-01" @default.
• W3106156541 modified "2023-10-18" @default.
• W3106156541 title "Adversarial Augmentation Policy Search for Domain and Cross-Lingual Generalization in Reading Comprehension" @default.
• W3106156541 cites W1502922572 @default.
• W3106156541 cites W1995945562 @default.
• W3106156541 cites W2016522586 @default.
• W3106156541 cites W2131241448 @default.
• W3106156541 cites W2155027007 @default.
• W3106156541 cites W2250539671 @default.
• W3106156541 cites W2557764419 @default.
• W3106156541 cites W2740132093 @default.
• W3106156541 cites W2798665661 @default.
• W3106156541 cites W2799194071 @default.
• W3106156541 cites W2890152674 @default.
• W3106156541 cites W2891017665 @default.
• W3106156541 cites W2946822591 @default.
• W3106156541 cites W2949128310 @default.
• W3106156541 cites W2949736877 @default.
• W3106156541 cites W2951665052 @default.
• W3106156541 cites W2951873305 @default.
• W3106156541 cites W2962713901 @default.
• W3106156541 cites W2962718684 @default.
• W3106156541 cites W2963126845 @default.
• W3106156541 cites W2963207607 @default.
• W3106156541 cites W2963323070 @default.
• W3106156541 cites W2963339397 @default.
• W3106156541 cites W2963341956 @default.
• W3106156541 cites W2963430447 @default.
• W3106156541 cites W2963542100 @default.
• W3106156541 cites W2963564796 @default.
• W3106156541 cites W2963661177 @default.
• W3106156541 cites W2963729324 @default.
• W3106156541 cites W2963748441 @default.
• W3106156541 cites W2963783970 @default.
• W3106156541 cites W2963903950 @default.
• W3106156541 cites W2963969878 @default.
• W3106156541 cites W2964159205 @default.
• W3106156541 cites W2964207259 @default.
• W3106156541 cites W2965373594 @default.
• W3106156541 cites W2970295111 @default.
• W3106156541 cites W2970418174 @default.
• W3106156541 cites W2970928735 @default.
• W3106156541 cites W2975761646 @default.
• W3106156541 cites W2978017171 @default.
• W3106156541 cites W2984402309 @default.
• W3106156541 cites W2985469912 @default.
• W3106156541 cites W2988421999 @default.
• W3106156541 cites W2997636389 @default.
• W3106156541 cites W3034469191 @default.
• W3106156541 cites W3035497479 @default.
• W3106156541 cites W3035743198 @default.
• W3106156541 cites W3093419064 @default.
• W3106156541 doi "https://doi.org/10.18653/v1/2020.findings-emnlp.333" @default.
• W3106156541 hasPublicationYear "2020" @default.
• W3106156541 type Work @default.
• W3106156541 sameAs 3106156541 @default.
• W3106156541 citedByCount "6" @default.
• W3106156541 countsByYear W31061565412020 @default.
• W3106156541 countsByYear W31061565412021 @default.
• W3106156541 countsByYear W31061565412022 @default.
• W3106156541 countsByYear W31061565412023 @default.
• W3106156541 crossrefType "proceedings-article" @default.
• W3106156541 hasAuthorship W3106156541A5001987532 @default.
• W3106156541 hasAuthorship W3106156541A5048025332 @default.
• W3106156541 hasBestOaLocation W31061565411 @default.
• W3106156541 hasConcept C134306372 @default.
• W3106156541 hasConcept C138885662 @default.
• W3106156541 hasConcept C154945302 @default.
• W3106156541 hasConcept C177148314 @default.
• W3106156541 hasConcept C199360897 @default.
• W3106156541 hasConcept C204321447 @default.
• W3106156541 hasConcept C2778780117 @default.
• W3106156541 hasConcept C33923547 @default.
• W3106156541 hasConcept C36503486 @default.
• W3106156541 hasConcept C37736160 @default.
• W3106156541 hasConcept C41008148 @default.
• W3106156541 hasConcept C41895202 @default.
• W3106156541 hasConcept C511192102 @default.
• W3106156541 hasConcept C554936623 @default.
• W3106156541 hasConceptScore W3106156541C134306372 @default.
• W3106156541 hasConceptScore W3106156541C138885662 @default.
• W3106156541 hasConceptScore W3106156541C154945302 @default.
• W3106156541 hasConceptScore W3106156541C177148314 @default.
• W3106156541 hasConceptScore W3106156541C199360897 @default.
• W3106156541 hasConceptScore W3106156541C204321447 @default.
• W3106156541 hasConceptScore W3106156541C2778780117 @default.
• W3106156541 hasConceptScore W3106156541C33923547 @default.
• W3106156541 hasConceptScore W3106156541C36503486 @default.
• W3106156541 hasConceptScore W3106156541C37736160 @default.
• W3106156541 hasConceptScore W3106156541C41008148 @default.
• W3106156541 hasConceptScore W3106156541C41895202 @default.
• W3106156541 hasConceptScore W3106156541C511192102 @default.
• W3106156541 hasConceptScore W3106156541C554936623 @default.
• W3106156541 hasLocation W31061565411 @default.
• W3106156541 hasLocation W31061565412 @default.
• W3106156541 hasOpenAccess W3106156541 @default.

• next