Matches in SemOpenAlex for { <https://semopenalex.org/work/W3016610080> ?p ?o ?g. }
- W3016610080 abstract "Lack of specialized data makes building a multi-domain neural machine translation tool challenging. Although emerging literature dealing with low resource languages starts to show promising results, most state-of-the-art models used millions of sentences. Today, the majority of multi-domain adaptation techniques are based on complex and sophisticated architectures that are not adapted for real-world applications. So far, no scalable method is performing better than the simple yet effective mixed-finetuning, i.e finetuning a generic model with a mix of all specialized data and generic data. In this paper, we propose a new training pipeline where knowledge distillation and multiple specialized teachers allow us to efficiently finetune a model without adding new costs at inference time. Our experiments demonstrated that our training pipeline allows improving the performance of multi-domain translation over finetuning in configurations with 2, 3, and 4 domains by up to 2 points in BLEU." @default.
- W3016610080 created "2020-04-24" @default.
- W3016610080 creator A5046630430 @default.
- W3016610080 creator A5054565289 @default.
- W3016610080 date "2020-04-15" @default.
- W3016610080 modified "2023-09-23" @default.
- W3016610080 title "Building a Multi-domain Neural Machine Translation Model using Knowledge Distillation" @default.
- W3016610080 cites W1821462560 @default.
- W3016610080 cites W1902237438 @default.
- W3016610080 cites W2064675550 @default.
- W3016610080 cites W2157331557 @default.
- W3016610080 cites W2542835211 @default.
- W3016610080 cites W2567571499 @default.
- W3016610080 cites W2574872930 @default.
- W3016610080 cites W2581863816 @default.
- W3016610080 cites W2586087033 @default.
- W3016610080 cites W2587694128 @default.
- W3016610080 cites W2750588180 @default.
- W3016610080 cites W2756978580 @default.
- W3016610080 cites W2784121710 @default.
- W3016610080 cites W2802153702 @default.
- W3016610080 cites W2805394970 @default.
- W3016610080 cites W2892244498 @default.
- W3016610080 cites W2905933322 @default.
- W3016610080 cites W2948210185 @default.
- W3016610080 cites W2949888546 @default.
- W3016610080 cites W2952444318 @default.
- W3016610080 cites W2963341956 @default.
- W3016610080 cites W2963366389 @default.
- W3016610080 cites W2963403868 @default.
- W3016610080 cites W2963532001 @default.
- W3016610080 cites W2963736842 @default.
- W3016610080 cites W2964308564 @default.
- W3016610080 cites W2965373594 @default.
- W3016610080 hasPublicationYear "2020" @default.
- W3016610080 type Work @default.
- W3016610080 sameAs 3016610080 @default.
- W3016610080 citedByCount "5" @default.
- W3016610080 countsByYear W30166100802020 @default.
- W3016610080 countsByYear W30166100802021 @default.
- W3016610080 crossrefType "posted-content" @default.
- W3016610080 hasAuthorship W3016610080A5046630430 @default.
- W3016610080 hasAuthorship W3016610080A5054565289 @default.
- W3016610080 hasConcept C104317684 @default.
- W3016610080 hasConcept C105580179 @default.
- W3016610080 hasConcept C111472728 @default.
- W3016610080 hasConcept C119857082 @default.
- W3016610080 hasConcept C120665830 @default.
- W3016610080 hasConcept C121332964 @default.
- W3016610080 hasConcept C134306372 @default.
- W3016610080 hasConcept C138885662 @default.
- W3016610080 hasConcept C139807058 @default.
- W3016610080 hasConcept C149364088 @default.
- W3016610080 hasConcept C154945302 @default.
- W3016610080 hasConcept C178790620 @default.
- W3016610080 hasConcept C185592680 @default.
- W3016610080 hasConcept C199360897 @default.
- W3016610080 hasConcept C203005215 @default.
- W3016610080 hasConcept C204030448 @default.
- W3016610080 hasConcept C207685749 @default.
- W3016610080 hasConcept C2776214188 @default.
- W3016610080 hasConcept C2776434776 @default.
- W3016610080 hasConcept C2780586882 @default.
- W3016610080 hasConcept C33923547 @default.
- W3016610080 hasConcept C36503486 @default.
- W3016610080 hasConcept C41008148 @default.
- W3016610080 hasConcept C43521106 @default.
- W3016610080 hasConcept C48044578 @default.
- W3016610080 hasConcept C55493867 @default.
- W3016610080 hasConcept C77088390 @default.
- W3016610080 hasConcept C95623464 @default.
- W3016610080 hasConceptScore W3016610080C104317684 @default.
- W3016610080 hasConceptScore W3016610080C105580179 @default.
- W3016610080 hasConceptScore W3016610080C111472728 @default.
- W3016610080 hasConceptScore W3016610080C119857082 @default.
- W3016610080 hasConceptScore W3016610080C120665830 @default.
- W3016610080 hasConceptScore W3016610080C121332964 @default.
- W3016610080 hasConceptScore W3016610080C134306372 @default.
- W3016610080 hasConceptScore W3016610080C138885662 @default.
- W3016610080 hasConceptScore W3016610080C139807058 @default.
- W3016610080 hasConceptScore W3016610080C149364088 @default.
- W3016610080 hasConceptScore W3016610080C154945302 @default.
- W3016610080 hasConceptScore W3016610080C178790620 @default.
- W3016610080 hasConceptScore W3016610080C185592680 @default.
- W3016610080 hasConceptScore W3016610080C199360897 @default.
- W3016610080 hasConceptScore W3016610080C203005215 @default.
- W3016610080 hasConceptScore W3016610080C204030448 @default.
- W3016610080 hasConceptScore W3016610080C207685749 @default.
- W3016610080 hasConceptScore W3016610080C2776214188 @default.
- W3016610080 hasConceptScore W3016610080C2776434776 @default.
- W3016610080 hasConceptScore W3016610080C2780586882 @default.
- W3016610080 hasConceptScore W3016610080C33923547 @default.
- W3016610080 hasConceptScore W3016610080C36503486 @default.
- W3016610080 hasConceptScore W3016610080C41008148 @default.
- W3016610080 hasConceptScore W3016610080C43521106 @default.
- W3016610080 hasConceptScore W3016610080C48044578 @default.
- W3016610080 hasConceptScore W3016610080C55493867 @default.
- W3016610080 hasConceptScore W3016610080C77088390 @default.
- W3016610080 hasConceptScore W3016610080C95623464 @default.
- W3016610080 hasLocation W30166100801 @default.