Matches in SemOpenAlex for { <https://semopenalex.org/work/W3200680973> ?p ?o ?g. }
- W3200680973 abstract "Task-adaptive pre-training (TAPT) and Self-training (ST) have emerged as the major semi-supervised approaches to improve natural language understanding (NLU) tasks with massive amount of unlabeled data. However, it's unclear whether they learn similar representations or they can be effectively combined. In this paper, we show that TAPT and ST can be complementary with simple TFS protocol by following TAPT -> Finetuning -> Self-training (TFS) process. Experimental results show that TFS protocol can effectively utilize unlabeled data to achieve strong combined gains consistently across six datasets covering sentiment classification, paraphrase identification, natural language inference, named entity recognition and dialogue slot classification. We investigate various semi-supervised settings and consistently show that gains from TAPT and ST can be strongly additive by following TFS procedure. We hope that TFS could serve as an important semi-supervised baseline for future NLP studies." @default.
- W3200680973 created "2021-09-27" @default.
- W3200680973 creator A5041666153 @default.
- W3200680973 creator A5047709762 @default.
- W3200680973 creator A5048500970 @default.
- W3200680973 creator A5063070498 @default.
- W3200680973 date "2021-09-14" @default.
- W3200680973 modified "2023-09-25" @default.
- W3200680973 title "Task-adaptive Pre-training and Self-training are Complementary for Natural Language Understanding" @default.
- W3200680973 cites W1821462560 @default.
- W3200680973 cites W1990334093 @default.
- W3200680973 cites W2111316763 @default.
- W3200680973 cites W2117539524 @default.
- W3200680973 cites W2144578941 @default.
- W3200680973 cites W2251939518 @default.
- W3200680973 cites W2911489562 @default.
- W3200680973 cites W2963310665 @default.
- W3200680973 cites W2963341956 @default.
- W3200680973 cites W2963748441 @default.
- W3200680973 cites W2963846996 @default.
- W3200680973 cites W2965373594 @default.
- W3200680973 cites W2990704537 @default.
- W3200680973 cites W2995746049 @default.
- W3200680973 cites W3034238904 @default.
- W3200680973 cites W3034978746 @default.
- W3200680973 cites W3035160371 @default.
- W3200680973 cites W3037311587 @default.
- W3200680973 cites W3082274269 @default.
- W3200680973 cites W3100110884 @default.
- W3200680973 cites W3100859887 @default.
- W3200680973 cites W3105724776 @default.
- W3200680973 cites W3120044914 @default.
- W3200680973 cites W3125853266 @default.
- W3200680973 cites W3168921237 @default.
- W3200680973 cites W9634341 @default.
- W3200680973 cites W3126359574 @default.
- W3200680973 doi "https://doi.org/10.48550/arxiv.2109.06466" @default.
- W3200680973 hasPublicationYear "2021" @default.
- W3200680973 type Work @default.
- W3200680973 sameAs 3200680973 @default.
- W3200680973 citedByCount "0" @default.
- W3200680973 crossrefType "posted-content" @default.
- W3200680973 hasAuthorship W3200680973A5041666153 @default.
- W3200680973 hasAuthorship W3200680973A5047709762 @default.
- W3200680973 hasAuthorship W3200680973A5048500970 @default.
- W3200680973 hasAuthorship W3200680973A5063070498 @default.
- W3200680973 hasBestOaLocation W32006809731 @default.
- W3200680973 hasConcept C111368507 @default.
- W3200680973 hasConcept C116834253 @default.
- W3200680973 hasConcept C119857082 @default.
- W3200680973 hasConcept C12725497 @default.
- W3200680973 hasConcept C127313418 @default.
- W3200680973 hasConcept C142724271 @default.
- W3200680973 hasConcept C154945302 @default.
- W3200680973 hasConcept C162324750 @default.
- W3200680973 hasConcept C187736073 @default.
- W3200680973 hasConcept C195324797 @default.
- W3200680973 hasConcept C204321447 @default.
- W3200680973 hasConcept C204787440 @default.
- W3200680973 hasConcept C2776214188 @default.
- W3200680973 hasConcept C2779439875 @default.
- W3200680973 hasConcept C2780385302 @default.
- W3200680973 hasConcept C2780451532 @default.
- W3200680973 hasConcept C2780922921 @default.
- W3200680973 hasConcept C41008148 @default.
- W3200680973 hasConcept C59822182 @default.
- W3200680973 hasConcept C71924100 @default.
- W3200680973 hasConcept C86803240 @default.
- W3200680973 hasConceptScore W3200680973C111368507 @default.
- W3200680973 hasConceptScore W3200680973C116834253 @default.
- W3200680973 hasConceptScore W3200680973C119857082 @default.
- W3200680973 hasConceptScore W3200680973C12725497 @default.
- W3200680973 hasConceptScore W3200680973C127313418 @default.
- W3200680973 hasConceptScore W3200680973C142724271 @default.
- W3200680973 hasConceptScore W3200680973C154945302 @default.
- W3200680973 hasConceptScore W3200680973C162324750 @default.
- W3200680973 hasConceptScore W3200680973C187736073 @default.
- W3200680973 hasConceptScore W3200680973C195324797 @default.
- W3200680973 hasConceptScore W3200680973C204321447 @default.
- W3200680973 hasConceptScore W3200680973C204787440 @default.
- W3200680973 hasConceptScore W3200680973C2776214188 @default.
- W3200680973 hasConceptScore W3200680973C2779439875 @default.
- W3200680973 hasConceptScore W3200680973C2780385302 @default.
- W3200680973 hasConceptScore W3200680973C2780451532 @default.
- W3200680973 hasConceptScore W3200680973C2780922921 @default.
- W3200680973 hasConceptScore W3200680973C41008148 @default.
- W3200680973 hasConceptScore W3200680973C59822182 @default.
- W3200680973 hasConceptScore W3200680973C71924100 @default.
- W3200680973 hasConceptScore W3200680973C86803240 @default.
- W3200680973 hasLocation W32006809731 @default.
- W3200680973 hasOpenAccess W3200680973 @default.
- W3200680973 hasPrimaryLocation W32006809731 @default.
- W3200680973 hasRelatedWork W1806995473 @default.
- W3200680973 hasRelatedWork W2038246283 @default.
- W3200680973 hasRelatedWork W2295682796 @default.
- W3200680973 hasRelatedWork W2587329402 @default.
- W3200680973 hasRelatedWork W2602143361 @default.
- W3200680973 hasRelatedWork W2612059125 @default.
- W3200680973 hasRelatedWork W2761417937 @default.
- W3200680973 hasRelatedWork W2977842567 @default.