FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3101597781> ?p ?o ?g. }

• W3101597781 abstract "Current state-of-the-art text generators build on powerful language models such as GPT-2, achieving impressive performance. However, to avoid degenerate text, they require sampling from a modified softmax, via temperature parameters or ad-hoc truncation techniques, as in top-k or nucleus sampling. This creates a mismatch between training and testing conditions. In this paper, we use the recently introduced entmax transformation to train and sample from a natively sparse language model, avoiding this mismatch. The result is a text generator with favorable performance in terms of fluency and consistency, fewer repetitions, and n-gram diversity closer to human text. In order to evaluate our model, we propose three new metrics for comparing sparse or truncated distributions: 𝜖-perplexity, sparsemax score, and Jensen-Shannon divergence. Human-evaluated experiments in story completion and dialogue generation show that entmax sampling leads to more engaging and coherent stories and conversations." @default.
• W3101597781 created "2020-11-23" @default.
• W3101597781 creator A5045450120 @default.
• W3101597781 creator A5051693368 @default.
• W3101597781 creator A5081932185 @default.
• W3101597781 date "2020-01-01" @default.
• W3101597781 modified "2023-10-17" @default.
• W3101597781 title "Sparse Text Generation" @default.
• W3101597781 cites W1566289585 @default.
• W3101597781 cites W1912497050 @default.
• W3101597781 cites W1983874169 @default.
• W3101597781 cites W2058373514 @default.
• W3101597781 cites W2096765209 @default.
• W3101597781 cites W2101105183 @default.
• W3101597781 cites W2130942839 @default.
• W3101597781 cites W2158195707 @default.
• W3101597781 cites W2167655509 @default.
• W3101597781 cites W2168029153 @default.
• W3101597781 cites W2515695833 @default.
• W3101597781 cites W2557436004 @default.
• W3101597781 cites W2581637843 @default.
• W3101597781 cites W2788277448 @default.
• W3101597781 cites W2898658996 @default.
• W3101597781 cites W2914204778 @default.
• W3101597781 cites W2950858167 @default.
• W3101597781 cites W2962784628 @default.
• W3101597781 cites W2962788902 @default.
• W3101597781 cites W2962839844 @default.
• W3101597781 cites W2962965405 @default.
• W3101597781 cites W2963096510 @default.
• W3101597781 cites W2963123301 @default.
• W3101597781 cites W2963167310 @default.
• W3101597781 cites W2963206148 @default.
• W3101597781 cites W2963283805 @default.
• W3101597781 cites W2963403868 @default.
• W3101597781 cites W2963456134 @default.
• W3101597781 cites W2963494889 @default.
• W3101597781 cites W2963825865 @default.
• W3101597781 cites W2964110616 @default.
• W3101597781 cites W2964121744 @default.
• W3101597781 cites W2964268978 @default.
• W3101597781 cites W2964308564 @default.
• W3101597781 cites W2971883198 @default.
• W3101597781 cites W2988647680 @default.
• W3101597781 cites W2995404354 @default.
• W3101597781 cites W2996287690 @default.
• W3101597781 cites W3035068109 @default.
• W3101597781 cites W854541894 @default.
• W3101597781 doi "https://doi.org/10.18653/v1/2020.emnlp-main.348" @default.
• W3101597781 hasPublicationYear "2020" @default.
• W3101597781 type Work @default.
• W3101597781 sameAs 3101597781 @default.
• W3101597781 citedByCount "20" @default.
• W3101597781 countsByYear W31015977812020 @default.
• W3101597781 countsByYear W31015977812021 @default.
• W3101597781 countsByYear W31015977812022 @default.
• W3101597781 countsByYear W31015977812023 @default.
• W3101597781 crossrefType "proceedings-article" @default.
• W3101597781 hasAuthorship W3101597781A5045450120 @default.
• W3101597781 hasAuthorship W3101597781A5051693368 @default.
• W3101597781 hasAuthorship W3101597781A5081932185 @default.
• W3101597781 hasBestOaLocation W31015977811 @default.
• W3101597781 hasConcept C100279451 @default.
• W3101597781 hasConcept C106131492 @default.
• W3101597781 hasConcept C121332964 @default.
• W3101597781 hasConcept C137293760 @default.
• W3101597781 hasConcept C138885662 @default.
• W3101597781 hasConcept C140779682 @default.
• W3101597781 hasConcept C145420912 @default.
• W3101597781 hasConcept C154945302 @default.
• W3101597781 hasConcept C163258240 @default.
• W3101597781 hasConcept C188441871 @default.
• W3101597781 hasConcept C204321447 @default.
• W3101597781 hasConcept C207390915 @default.
• W3101597781 hasConcept C2776436953 @default.
• W3101597781 hasConcept C2777413886 @default.
• W3101597781 hasConcept C2780992000 @default.
• W3101597781 hasConcept C31972630 @default.
• W3101597781 hasConcept C33923547 @default.
• W3101597781 hasConcept C41008148 @default.
• W3101597781 hasConcept C41895202 @default.
• W3101597781 hasConcept C50644808 @default.
• W3101597781 hasConcept C62520636 @default.
• W3101597781 hasConceptScore W3101597781C100279451 @default.
• W3101597781 hasConceptScore W3101597781C106131492 @default.
• W3101597781 hasConceptScore W3101597781C121332964 @default.
• W3101597781 hasConceptScore W3101597781C137293760 @default.
• W3101597781 hasConceptScore W3101597781C138885662 @default.
• W3101597781 hasConceptScore W3101597781C140779682 @default.
• W3101597781 hasConceptScore W3101597781C145420912 @default.
• W3101597781 hasConceptScore W3101597781C154945302 @default.
• W3101597781 hasConceptScore W3101597781C163258240 @default.
• W3101597781 hasConceptScore W3101597781C188441871 @default.
• W3101597781 hasConceptScore W3101597781C204321447 @default.
• W3101597781 hasConceptScore W3101597781C207390915 @default.
• W3101597781 hasConceptScore W3101597781C2776436953 @default.
• W3101597781 hasConceptScore W3101597781C2777413886 @default.
• W3101597781 hasConceptScore W3101597781C2780992000 @default.
• W3101597781 hasConceptScore W3101597781C31972630 @default.
• W3101597781 hasConceptScore W3101597781C33923547 @default.

• next