Matches in SemOpenAlex for { <https://semopenalex.org/work/W2986867746> ?p ?o ?g. }
Showing items 1 to 66 of
66
with 100 items per page.
- W2986867746 abstract "Leveraging external knowledge to enhance conversational models has become a popular research area in recent years. Compared to vanilla generative models, the knowledge-grounded models may produce more informative and engaging responses. Although various approaches have been proposed in the past, how to effectively incorporate knowledge remains an open research question. It is unclear how much external knowledge should be retrieved and what is the optimal way to enhance the conversational model, trading off between relevant information and noise. Therefore, in this paper, we aim to bridge the gap by first extensively evaluating various types of state-of-the-art knowledge-grounded conversational models, including recurrent neural network based, memory networks based, and Transformer based models. We demonstrate empirically that those conversational models can only be enhanced with the right amount of external knowledge. To effectively leverage information originated from external knowledge, we propose a novel Transformer with Expanded Decoder (Transformer-ED or TED for short), which can automatically tune the weights for different sources of evidence when generating responses. Our experiments show that our proposed model outperforms state-of-the-art models in terms of both quality and diversity." @default.
- W2986867746 created "2019-11-22" @default.
- W2986867746 creator A5015061573 @default.
- W2986867746 creator A5035367145 @default.
- W2986867746 date "2019-11-03" @default.
- W2986867746 modified "2023-09-30" @default.
- W2986867746 title "Enhancing Conversational Dialogue Models with Grounded Knowledge" @default.
- W2986867746 cites W1993378086 @default.
- W2986867746 cites W2133459682 @default.
- W2986867746 cites W2734443755 @default.
- W2986867746 cites W2741363662 @default.
- W2986867746 cites W2798392716 @default.
- W2986867746 cites W2885421725 @default.
- W2986867746 cites W2907833097 @default.
- W2986867746 cites W2963206148 @default.
- W2986867746 cites W2963789888 @default.
- W2986867746 cites W2964110616 @default.
- W2986867746 doi "https://doi.org/10.1145/3357384.3357889" @default.
- W2986867746 hasPublicationYear "2019" @default.
- W2986867746 type Work @default.
- W2986867746 sameAs 2986867746 @default.
- W2986867746 citedByCount "17" @default.
- W2986867746 countsByYear W29868677462020 @default.
- W2986867746 countsByYear W29868677462021 @default.
- W2986867746 countsByYear W29868677462022 @default.
- W2986867746 countsByYear W29868677462023 @default.
- W2986867746 crossrefType "proceedings-article" @default.
- W2986867746 hasAuthorship W2986867746A5015061573 @default.
- W2986867746 hasAuthorship W2986867746A5035367145 @default.
- W2986867746 hasConcept C119857082 @default.
- W2986867746 hasConcept C121332964 @default.
- W2986867746 hasConcept C153083717 @default.
- W2986867746 hasConcept C154945302 @default.
- W2986867746 hasConcept C165801399 @default.
- W2986867746 hasConcept C39890363 @default.
- W2986867746 hasConcept C41008148 @default.
- W2986867746 hasConcept C50644808 @default.
- W2986867746 hasConcept C62520636 @default.
- W2986867746 hasConcept C66322947 @default.
- W2986867746 hasConceptScore W2986867746C119857082 @default.
- W2986867746 hasConceptScore W2986867746C121332964 @default.
- W2986867746 hasConceptScore W2986867746C153083717 @default.
- W2986867746 hasConceptScore W2986867746C154945302 @default.
- W2986867746 hasConceptScore W2986867746C165801399 @default.
- W2986867746 hasConceptScore W2986867746C39890363 @default.
- W2986867746 hasConceptScore W2986867746C41008148 @default.
- W2986867746 hasConceptScore W2986867746C50644808 @default.
- W2986867746 hasConceptScore W2986867746C62520636 @default.
- W2986867746 hasConceptScore W2986867746C66322947 @default.
- W2986867746 hasLocation W29868677461 @default.
- W2986867746 hasOpenAccess W2986867746 @default.
- W2986867746 hasPrimaryLocation W29868677461 @default.
- W2986867746 hasRelatedWork W2961085424 @default.
- W2986867746 hasRelatedWork W3046775127 @default.
- W2986867746 hasRelatedWork W3170094116 @default.
- W2986867746 hasRelatedWork W3212218375 @default.
- W2986867746 hasRelatedWork W4285260836 @default.
- W2986867746 hasRelatedWork W4286629047 @default.
- W2986867746 hasRelatedWork W4306321456 @default.
- W2986867746 hasRelatedWork W4306674287 @default.
- W2986867746 hasRelatedWork W4321153836 @default.
- W2986867746 hasRelatedWork W4224009465 @default.
- W2986867746 isParatext "false" @default.
- W2986867746 isRetracted "false" @default.
- W2986867746 magId "2986867746" @default.
- W2986867746 workType "article" @default.