FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3039602870> ?p ?o ?g. }

• W3039602870 abstract "Prototype-driven text generation uses non-parametric models that first choose from a library of sentence and then modify the prototype to generate the output text. While effective, these methods are inefficient at test time as a result of needing to store and index the entire training corpus. Further, existing methods often require heuristics to identify which prototypes to reference at training time. In this paper, we propose a novel generative model that automatically learns a sparse prototype support set that, nonetheless, achieves strong language modeling performance. This is achieved by (1) imposing a sparsity-inducing prior on the prototype selection distribution, and (2) utilizing amortized variational inference to learn a prototype retrieval function. In experiments, our model outperforms previous prototype-driven language models while achieving up to a 1000x memory reduction, as well as a 1000x speed-up at test time. More interestingly, we show that the learned prototypes are able to capture semantics and syntax at different granularity as we vary the sparsity of prototype selection, and that certain sentence attributes can be controlled by specifying the prototype for generation." @default.
• W3039602870 created "2020-07-10" @default.
• W3039602870 creator A5015879697 @default.
• W3039602870 creator A5017455302 @default.
• W3039602870 creator A5068811427 @default.
• W3039602870 date "2020-06-29" @default.
• W3039602870 modified "2023-09-23" @default.
• W3039602870 title "Learning Sparse Prototypes for Text Generation" @default.
• W3039602870 cites W1518951372 @default.
• W3039602870 cites W1522301498 @default.
• W3039602870 cites W1663973292 @default.
• W3039602870 cites W179875071 @default.
• W3039602870 cites W1861492603 @default.
• W3039602870 cites W2064675550 @default.
• W3039602870 cites W2119717200 @default.
• W3039602870 cites W2132339004 @default.
• W3039602870 cites W2142112143 @default.
• W3039602870 cites W2166851633 @default.
• W3039602870 cites W2250539671 @default.
• W3039602870 cites W2402268235 @default.
• W3039602870 cites W2549835527 @default.
• W3039602870 cites W2554915555 @default.
• W3039602870 cites W2587284713 @default.
• W3039602870 cites W2788330850 @default.
• W3039602870 cites W2885421725 @default.
• W3039602870 cites W2889467844 @default.
• W3039602870 cites W2890397703 @default.
• W3039602870 cites W2933138175 @default.
• W3039602870 cites W2951004968 @default.
• W3039602870 cites W2962832505 @default.
• W3039602870 cites W2962965405 @default.
• W3039602870 cites W2963018920 @default.
• W3039602870 cites W2963081964 @default.
• W3039602870 cites W2963088785 @default.
• W3039602870 cites W2963149635 @default.
• W3039602870 cites W2963223306 @default.
• W3039602870 cites W2963275229 @default.
• W3039602870 cites W2963341956 @default.
• W3039602870 cites W2963371754 @default.
• W3039602870 cites W2963403868 @default.
• W3039602870 cites W2963667126 @default.
• W3039602870 cites W2963965612 @default.
• W3039602870 cites W2964000524 @default.
• W3039602870 cites W2964110616 @default.
• W3039602870 cites W2964308564 @default.
• W3039602870 cites W2965373594 @default.
• W3039602870 cites W2969442125 @default.
• W3039602870 cites W2970134931 @default.
• W3039602870 cites W2970476646 @default.
• W3039602870 cites W2970597249 @default.
• W3039602870 cites W2971008823 @default.
• W3039602870 cites W2980282514 @default.
• W3039602870 cites W2995154514 @default.
• W3039602870 cites W3037109418 @default.
• W3039602870 cites W3090145300 @default.
• W3039602870 cites W68733909 @default.
• W3039602870 hasPublicationYear "2020" @default.
• W3039602870 type Work @default.
• W3039602870 sameAs 3039602870 @default.
• W3039602870 citedByCount "3" @default.
• W3039602870 countsByYear W30396028702021 @default.
• W3039602870 crossrefType "posted-content" @default.
• W3039602870 hasAuthorship W3039602870A5015879697 @default.
• W3039602870 hasAuthorship W3039602870A5017455302 @default.
• W3039602870 hasAuthorship W3039602870A5068811427 @default.
• W3039602870 hasConcept C107673813 @default.
• W3039602870 hasConcept C111919701 @default.
• W3039602870 hasConcept C119857082 @default.
• W3039602870 hasConcept C127705205 @default.
• W3039602870 hasConcept C13280743 @default.
• W3039602870 hasConcept C137293760 @default.
• W3039602870 hasConcept C14036430 @default.
• W3039602870 hasConcept C154945302 @default.
• W3039602870 hasConcept C160234255 @default.
• W3039602870 hasConcept C167966045 @default.
• W3039602870 hasConcept C169903167 @default.
• W3039602870 hasConcept C177264268 @default.
• W3039602870 hasConcept C185798385 @default.
• W3039602870 hasConcept C199360897 @default.
• W3039602870 hasConcept C205649164 @default.
• W3039602870 hasConcept C2776214188 @default.
• W3039602870 hasConcept C2777530160 @default.
• W3039602870 hasConcept C39890363 @default.
• W3039602870 hasConcept C41008148 @default.
• W3039602870 hasConcept C60048249 @default.
• W3039602870 hasConcept C78458016 @default.
• W3039602870 hasConcept C81917197 @default.
• W3039602870 hasConcept C86803240 @default.
• W3039602870 hasConceptScore W3039602870C107673813 @default.
• W3039602870 hasConceptScore W3039602870C111919701 @default.
• W3039602870 hasConceptScore W3039602870C119857082 @default.
• W3039602870 hasConceptScore W3039602870C127705205 @default.
• W3039602870 hasConceptScore W3039602870C13280743 @default.
• W3039602870 hasConceptScore W3039602870C137293760 @default.
• W3039602870 hasConceptScore W3039602870C14036430 @default.
• W3039602870 hasConceptScore W3039602870C154945302 @default.
• W3039602870 hasConceptScore W3039602870C160234255 @default.
• W3039602870 hasConceptScore W3039602870C167966045 @default.
• W3039602870 hasConceptScore W3039602870C169903167 @default.
• W3039602870 hasConceptScore W3039602870C177264268 @default.

• next