Matches in SemOpenAlex for { <https://semopenalex.org/work/W3118509208> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3118509208 abstract "Transformers have brought a remarkable improvement in the performance of neural machine translation (NMT) systems but they could be surprisingly vulnerable to noise. In this work, we try to investigate how noise breaks Transformers and if there exist solutions to deal with such issues. There is a large body of work in the NMT literature on analyzing the behavior of conventional models for the problem of noise but Transformers are relatively understudied in this context. Motivated by this, we introduce a novel data-driven technique called Target Augmented Fine-tuning (TAFT) to incorporate noise during training. This idea is comparable to the well-known fine-tuning strategy. Moreover, we propose two other novel extensions to the original Transformer: Controlled Denoising (CD) and Dual-Channel Decoding (DCD), that modify the neural architecture as well as the training process to handle noise. One important characteristic of our techniques is that they only impact the training phase and do not impose any overhead at inference time. We evaluated our techniques to translate the English–German pair in both directions and observed that our models have a higher tolerance to noise. More specifically, they perform with no deterioration where up to 10% of entire test words are infected by noise." @default.
- W3118509208 created "2021-01-18" @default.
- W3118509208 creator A5003270434 @default.
- W3118509208 creator A5066887020 @default.
- W3118509208 creator A5076656061 @default.
- W3118509208 date "2021-01-01" @default.
- W3118509208 modified "2023-09-27" @default.
- W3118509208 title "Revisiting Robust Neural Machine Translation: A Transformer Case Study" @default.
- W3118509208 cites W1522301498 @default.
- W3118509208 cites W1938755728 @default.
- W3118509208 cites W2101105183 @default.
- W3118509208 cites W2120615054 @default.
- W3118509208 cites W2130942839 @default.
- W3118509208 cites W2157331557 @default.
- W3118509208 cites W2552839021 @default.
- W3118509208 cites W2962784628 @default.
- W3118509208 cites W2963341956 @default.
- W3118509208 cites W2963403868 @default.
- W3118509208 cites W2963532001 @default.
- W3118509208 cites W2963661177 @default.
- W3118509208 cites W2963823140 @default.
- W3118509208 cites W2963919854 @default.
- W3118509208 cites W2964048171 @default.
- W3118509208 cites W2964132420 @default.
- W3118509208 cites W2964199361 @default.
- W3118509208 cites W2964247056 @default.
- W3118509208 cites W2970169061 @default.
- W3118509208 cites W2970558573 @default.
- W3118509208 cites W2970798168 @default.
- W3118509208 cites W2970926924 @default.
- W3118509208 cites W2971134989 @default.
- W3118509208 cites W2984051011 @default.
- W3118509208 cites W2984060780 @default.
- W3118509208 cites W3005724337 @default.
- W3118509208 cites W3035812575 @default.
- W3118509208 doi "https://doi.org/10.18653/v1/2021.findings-emnlp.323" @default.
- W3118509208 hasPublicationYear "2021" @default.
- W3118509208 type Work @default.
- W3118509208 sameAs 3118509208 @default.
- W3118509208 citedByCount "3" @default.
- W3118509208 countsByYear W31185092082021 @default.
- W3118509208 crossrefType "proceedings-article" @default.
- W3118509208 hasAuthorship W3118509208A5003270434 @default.
- W3118509208 hasAuthorship W3118509208A5066887020 @default.
- W3118509208 hasAuthorship W3118509208A5076656061 @default.
- W3118509208 hasBestOaLocation W31185092081 @default.
- W3118509208 hasConcept C113775141 @default.
- W3118509208 hasConcept C11413529 @default.
- W3118509208 hasConcept C115961682 @default.
- W3118509208 hasConcept C119599485 @default.
- W3118509208 hasConcept C119857082 @default.
- W3118509208 hasConcept C127413603 @default.
- W3118509208 hasConcept C154945302 @default.
- W3118509208 hasConcept C163294075 @default.
- W3118509208 hasConcept C165801399 @default.
- W3118509208 hasConcept C203005215 @default.
- W3118509208 hasConcept C2776214188 @default.
- W3118509208 hasConcept C41008148 @default.
- W3118509208 hasConcept C57273362 @default.
- W3118509208 hasConcept C66322947 @default.
- W3118509208 hasConcept C99498987 @default.
- W3118509208 hasConceptScore W3118509208C113775141 @default.
- W3118509208 hasConceptScore W3118509208C11413529 @default.
- W3118509208 hasConceptScore W3118509208C115961682 @default.
- W3118509208 hasConceptScore W3118509208C119599485 @default.
- W3118509208 hasConceptScore W3118509208C119857082 @default.
- W3118509208 hasConceptScore W3118509208C127413603 @default.
- W3118509208 hasConceptScore W3118509208C154945302 @default.
- W3118509208 hasConceptScore W3118509208C163294075 @default.
- W3118509208 hasConceptScore W3118509208C165801399 @default.
- W3118509208 hasConceptScore W3118509208C203005215 @default.
- W3118509208 hasConceptScore W3118509208C2776214188 @default.
- W3118509208 hasConceptScore W3118509208C41008148 @default.
- W3118509208 hasConceptScore W3118509208C57273362 @default.
- W3118509208 hasConceptScore W3118509208C66322947 @default.
- W3118509208 hasConceptScore W3118509208C99498987 @default.
- W3118509208 hasLocation W31185092081 @default.
- W3118509208 hasLocation W31185092082 @default.
- W3118509208 hasOpenAccess W3118509208 @default.
- W3118509208 hasPrimaryLocation W31185092081 @default.
- W3118509208 hasRelatedWork W1988327103 @default.
- W3118509208 hasRelatedWork W2833393231 @default.
- W3118509208 hasRelatedWork W2917049097 @default.
- W3118509208 hasRelatedWork W2947946877 @default.
- W3118509208 hasRelatedWork W2999635570 @default.
- W3118509208 hasRelatedWork W3107474891 @default.
- W3118509208 hasRelatedWork W3127943074 @default.
- W3118509208 hasRelatedWork W3206689743 @default.
- W3118509208 hasRelatedWork W3214236736 @default.
- W3118509208 hasRelatedWork W4303874710 @default.
- W3118509208 isParatext "false" @default.
- W3118509208 isRetracted "false" @default.
- W3118509208 magId "3118509208" @default.
- W3118509208 workType "article" @default.