FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3034328552> ?p ?o ?g. }

• W3034328552 abstract "Many NLP tasks such as tagging and machine reading comprehension are faced with the severe data imbalance issue: negative examples significantly outnumber positive examples, and the huge number of easy-negative examples overwhelms the training. The most commonly used cross entropy (CE) criteria is actually an accuracy-oriented objective, and thus creates a discrepancy between training and test: at training time, each training instance contributes equally to the objective function, while at test time F1 score concerns more about positive examples. In this paper, we propose to use dice loss in replacement of the standard cross-entropy objective for data-imbalanced NLP tasks. Dice loss is based on the Sørensen--Dice coefficient or Tversky index , which attaches similar importance to false positives and false negatives, and is more immune to the data-imbalance issue. To further alleviate the dominating influence from easy-negative examples in training, we propose to associate training examples with dynamically adjusted weights to deemphasize easy-negative examples. Theoretical analysis shows that this strategy narrows down the gap between the F1 score in evaluation and the dice loss in training. With the proposed training objective, we observe significant performance boost on a wide range of data imbalanced NLP tasks. Notably, we are able to achieve SOTA results on CTB5, CTB6 and UD1.4 for the part of speech tagging task; SOTA results on CoNLL03, OntoNotes5.0, MSRA and OntoNotes4.0 for the named entity recognition task; along with competitive results on the tasks of machine reading comprehension and paraphrase identification." @default.
• W3034328552 created "2020-06-19" @default.
• W3034328552 creator A5004882141 @default.
• W3034328552 creator A5014777126 @default.
• W3034328552 creator A5025603106 @default.
• W3034328552 creator A5044736298 @default.
• W3034328552 creator A5074612639 @default.
• W3034328552 creator A5077851706 @default.
• W3034328552 date "2020-01-01" @default.
• W3034328552 modified "2023-10-10" @default.
• W3034328552 title "Dice Loss for Data-imbalanced NLP Tasks" @default.
• W3034328552 cites W131533222 @default.
• W3034328552 cites W1536680647 @default.
• W3034328552 cites W1934410531 @default.
• W3034328552 cites W1987869189 @default.
• W3034328552 cites W1989684337 @default.
• W3034328552 cites W2020278455 @default.
• W3034328552 cites W2059975159 @default.
• W3034328552 cites W2063471322 @default.
• W3034328552 cites W2096204319 @default.
• W3034328552 cites W2096235960 @default.
• W3034328552 cites W2102605133 @default.
• W3034328552 cites W2103076621 @default.
• W3034328552 cites W2132984949 @default.
• W3034328552 cites W2144578941 @default.
• W3034328552 cites W2148143831 @default.
• W3034328552 cites W2153848201 @default.
• W3034328552 cites W2194775991 @default.
• W3034328552 cites W2252066972 @default.
• W3034328552 cites W2286855692 @default.
• W3034328552 cites W2296283641 @default.
• W3034328552 cites W2432481613 @default.
• W3034328552 cites W2516930406 @default.
• W3034328552 cites W2521709538 @default.
• W3034328552 cites W2551396370 @default.
• W3034328552 cites W2611775752 @default.
• W3034328552 cites W2734349601 @default.
• W3034328552 cites W2784195389 @default.
• W3034328552 cites W2792891771 @default.
• W3034328552 cites W2798305197 @default.
• W3034328552 cites W2809324505 @default.
• W3034328552 cites W2896649846 @default.
• W3034328552 cites W2903432333 @default.
• W3034328552 cites W2947925742 @default.
• W3034328552 cites W2951534261 @default.
• W3034328552 cites W2962721361 @default.
• W3034328552 cites W2962739339 @default.
• W3034328552 cites W2962902328 @default.
• W3034328552 cites W2962904552 @default.
• W3034328552 cites W2963081269 @default.
• W3034328552 cites W2963323070 @default.
• W3034328552 cites W2963336393 @default.
• W3034328552 cites W2963341956 @default.
• W3034328552 cites W2963351448 @default.
• W3034328552 cites W2963476860 @default.
• W3034328552 cites W2963564796 @default.
• W3034328552 cites W2963748441 @default.
• W3034328552 cites W2963933682 @default.
• W3034328552 cites W2969875432 @default.
• W3034328552 cites W2970597249 @default.
• W3034328552 cites W2970681607 @default.
• W3034328552 cites W2970745243 @default.
• W3034328552 cites W3035625205 @default.
• W3034328552 cites W639708223 @default.
• W3034328552 doi "https://doi.org/10.18653/v1/2020.acl-main.45" @default.
• W3034328552 hasPublicationYear "2020" @default.
• W3034328552 type Work @default.
• W3034328552 sameAs 3034328552 @default.
• W3034328552 citedByCount "119" @default.
• W3034328552 countsByYear W30343285522020 @default.
• W3034328552 countsByYear W30343285522021 @default.
• W3034328552 countsByYear W30343285522022 @default.
• W3034328552 countsByYear W30343285522023 @default.
• W3034328552 crossrefType "proceedings-article" @default.
• W3034328552 hasAuthorship W3034328552A5004882141 @default.
• W3034328552 hasAuthorship W3034328552A5014777126 @default.
• W3034328552 hasAuthorship W3034328552A5025603106 @default.
• W3034328552 hasAuthorship W3034328552A5044736298 @default.
• W3034328552 hasAuthorship W3034328552A5074612639 @default.
• W3034328552 hasAuthorship W3034328552A5077851706 @default.
• W3034328552 hasBestOaLocation W30343285522 @default.
• W3034328552 hasConcept C105795698 @default.
• W3034328552 hasConcept C119857082 @default.
• W3034328552 hasConcept C12267149 @default.
• W3034328552 hasConcept C154945302 @default.
• W3034328552 hasConcept C162324750 @default.
• W3034328552 hasConcept C167981619 @default.
• W3034328552 hasConcept C16910744 @default.
• W3034328552 hasConcept C187736073 @default.
• W3034328552 hasConcept C199360897 @default.
• W3034328552 hasConcept C204321447 @default.
• W3034328552 hasConcept C22029948 @default.
• W3034328552 hasConcept C2780451532 @default.
• W3034328552 hasConcept C2780922921 @default.
• W3034328552 hasConcept C33923547 @default.
• W3034328552 hasConcept C41008148 @default.
• W3034328552 hasConcept C51632099 @default.
• W3034328552 hasConcept C64869954 @default.
• W3034328552 hasConcept C9679016 @default.
• W3034328552 hasConceptScore W3034328552C105795698 @default.

• next