Matches in SemOpenAlex for { <https://semopenalex.org/work/W3137214344> ?p ?o ?g. }
- W3137214344 abstract "Currently used semantic parsing systems deployed in voice assistants can require weeks to train. Datasets for these models often receive small and frequent updates, data patches. Each patch requires training a new model. To reduce training time, one can fine-tune the previously trained model on each patch, but naive fine-tuning exhibits catastrophic forgetting - degradation of the model performance on the data not represented in the data patch. In this work, we propose a simple method that alleviates catastrophic forgetting and show that it is possible to match the performance of a model trained from scratch in less than 10% of a time via fine-tuning. The key to achieving this is supersampling and EWC regularization. We demonstrate the effectiveness of our method on multiple splits of the Facebook TOP and SNIPS datasets." @default.
- W3137214344 created "2021-03-29" @default.
- W3137214344 creator A5026934685 @default.
- W3137214344 creator A5030535530 @default.
- W3137214344 creator A5040387457 @default.
- W3137214344 creator A5071360545 @default.
- W3137214344 creator A5080273887 @default.
- W3137214344 date "2020-10-15" @default.
- W3137214344 modified "2023-09-27" @default.
- W3137214344 title "Update Frequently, Update Fast: Retraining Semantic Parsing Systems in a Fraction of Time" @default.
- W3137214344 cites W1682403713 @default.
- W3137214344 cites W2060277733 @default.
- W3137214344 cites W2141559645 @default.
- W3137214344 cites W2560647685 @default.
- W3137214344 cites W2737492962 @default.
- W3137214344 cites W2777054756 @default.
- W3137214344 cites W2788388592 @default.
- W3137214344 cites W2803609229 @default.
- W3137214344 cites W2892248135 @default.
- W3137214344 cites W2963026768 @default.
- W3137214344 cites W2963341956 @default.
- W3137214344 cites W2963403868 @default.
- W3137214344 cites W2963844079 @default.
- W3137214344 cites W2970172215 @default.
- W3137214344 cites W2971079754 @default.
- W3137214344 cites W3001279689 @default.
- W3137214344 cites W3002800333 @default.
- W3137214344 cites W3082274269 @default.
- W3137214344 cites W3094364292 @default.
- W3137214344 cites W3099351824 @default.
- W3137214344 cites W3100152912 @default.
- W3137214344 cites W3103469330 @default.
- W3137214344 hasPublicationYear "2020" @default.
- W3137214344 type Work @default.
- W3137214344 sameAs 3137214344 @default.
- W3137214344 citedByCount "4" @default.
- W3137214344 countsByYear W31372143442021 @default.
- W3137214344 crossrefType "posted-content" @default.
- W3137214344 hasAuthorship W3137214344A5026934685 @default.
- W3137214344 hasAuthorship W3137214344A5030535530 @default.
- W3137214344 hasAuthorship W3137214344A5040387457 @default.
- W3137214344 hasAuthorship W3137214344A5071360545 @default.
- W3137214344 hasAuthorship W3137214344A5080273887 @default.
- W3137214344 hasConcept C111472728 @default.
- W3137214344 hasConcept C111919701 @default.
- W3137214344 hasConcept C119857082 @default.
- W3137214344 hasConcept C138885662 @default.
- W3137214344 hasConcept C144133560 @default.
- W3137214344 hasConcept C154945302 @default.
- W3137214344 hasConcept C155202549 @default.
- W3137214344 hasConcept C186644900 @default.
- W3137214344 hasConcept C26517878 @default.
- W3137214344 hasConcept C2776135515 @default.
- W3137214344 hasConcept C2778712577 @default.
- W3137214344 hasConcept C2780586882 @default.
- W3137214344 hasConcept C2781235140 @default.
- W3137214344 hasConcept C38652104 @default.
- W3137214344 hasConcept C41008148 @default.
- W3137214344 hasConcept C41895202 @default.
- W3137214344 hasConcept C51632099 @default.
- W3137214344 hasConcept C7149132 @default.
- W3137214344 hasConceptScore W3137214344C111472728 @default.
- W3137214344 hasConceptScore W3137214344C111919701 @default.
- W3137214344 hasConceptScore W3137214344C119857082 @default.
- W3137214344 hasConceptScore W3137214344C138885662 @default.
- W3137214344 hasConceptScore W3137214344C144133560 @default.
- W3137214344 hasConceptScore W3137214344C154945302 @default.
- W3137214344 hasConceptScore W3137214344C155202549 @default.
- W3137214344 hasConceptScore W3137214344C186644900 @default.
- W3137214344 hasConceptScore W3137214344C26517878 @default.
- W3137214344 hasConceptScore W3137214344C2776135515 @default.
- W3137214344 hasConceptScore W3137214344C2778712577 @default.
- W3137214344 hasConceptScore W3137214344C2780586882 @default.
- W3137214344 hasConceptScore W3137214344C2781235140 @default.
- W3137214344 hasConceptScore W3137214344C38652104 @default.
- W3137214344 hasConceptScore W3137214344C41008148 @default.
- W3137214344 hasConceptScore W3137214344C41895202 @default.
- W3137214344 hasConceptScore W3137214344C51632099 @default.
- W3137214344 hasConceptScore W3137214344C7149132 @default.
- W3137214344 hasLocation W31372143441 @default.
- W3137214344 hasOpenAccess W3137214344 @default.
- W3137214344 hasPrimaryLocation W31372143441 @default.
- W3137214344 hasRelatedWork W2747206364 @default.
- W3137214344 hasRelatedWork W2755427716 @default.
- W3137214344 hasRelatedWork W2805965229 @default.
- W3137214344 hasRelatedWork W2884115684 @default.
- W3137214344 hasRelatedWork W2981864462 @default.
- W3137214344 hasRelatedWork W3014648906 @default.
- W3137214344 hasRelatedWork W3035204084 @default.
- W3137214344 hasRelatedWork W3046971235 @default.
- W3137214344 hasRelatedWork W3047670284 @default.
- W3137214344 hasRelatedWork W3092741951 @default.
- W3137214344 hasRelatedWork W3112928615 @default.
- W3137214344 hasRelatedWork W3115879670 @default.
- W3137214344 hasRelatedWork W3125127048 @default.
- W3137214344 hasRelatedWork W3127129507 @default.
- W3137214344 hasRelatedWork W3152401989 @default.
- W3137214344 hasRelatedWork W3161313759 @default.
- W3137214344 hasRelatedWork W3163086746 @default.
- W3137214344 hasRelatedWork W3166304536 @default.