Matches in SemOpenAlex for { <https://semopenalex.org/work/W4288104915> ?p ?o ?g. }
Showing items 1 to 67 of
67
with 100 items per page.
- W4288104915 abstract "The ability to understand and work with numbers (numeracy) is critical for many complex reasoning tasks. Currently, most NLP models treat numbers in text in the same way as other tokens---they embed them as distributed vectors. Is this enough to capture numeracy? We begin by investigating the numerical reasoning capabilities of a state-of-the-art question answering model on the DROP dataset. We find this model excels on questions that require numerical reasoning, i.e., it already captures numeracy. To understand how this capability emerges, we probe token embedding methods (e.g., BERT, GloVe) on synthetic list maximum, number decoding, and addition tasks. A surprising degree of numeracy is naturally present in standard embeddings. For example, GloVe and word2vec accurately encode magnitude for numbers up to 1,000. Furthermore, character-level embeddings are even more precise---ELMo captures numeracy the best for all pre-trained methods---but BERT, which uses sub-word units, is less exact." @default.
- W4288104915 created "2022-07-28" @default.
- W4288104915 creator A5005779128 @default.
- W4288104915 creator A5032740979 @default.
- W4288104915 creator A5035088083 @default.
- W4288104915 creator A5058353424 @default.
- W4288104915 creator A5077959671 @default.
- W4288104915 date "2019-09-17" @default.
- W4288104915 modified "2023-10-16" @default.
- W4288104915 title "Do NLP Models Know Numbers? Probing Numeracy in Embeddings" @default.
- W4288104915 doi "https://doi.org/10.48550/arxiv.1909.07940" @default.
- W4288104915 hasPublicationYear "2019" @default.
- W4288104915 type Work @default.
- W4288104915 citedByCount "0" @default.
- W4288104915 crossrefType "posted-content" @default.
- W4288104915 hasAuthorship W4288104915A5005779128 @default.
- W4288104915 hasAuthorship W4288104915A5032740979 @default.
- W4288104915 hasAuthorship W4288104915A5035088083 @default.
- W4288104915 hasAuthorship W4288104915A5058353424 @default.
- W4288104915 hasAuthorship W4288104915A5077959671 @default.
- W4288104915 hasBestOaLocation W42881049151 @default.
- W4288104915 hasConcept C154945302 @default.
- W4288104915 hasConcept C162324750 @default.
- W4288104915 hasConcept C204321447 @default.
- W4288104915 hasConcept C2524010 @default.
- W4288104915 hasConcept C2776461190 @default.
- W4288104915 hasConcept C33923547 @default.
- W4288104915 hasConcept C38652104 @default.
- W4288104915 hasConcept C41008148 @default.
- W4288104915 hasConcept C41608201 @default.
- W4288104915 hasConcept C48145219 @default.
- W4288104915 hasConcept C50522688 @default.
- W4288104915 hasConcept C53537400 @default.
- W4288104915 hasConcept C547764534 @default.
- W4288104915 hasConcept C80444323 @default.
- W4288104915 hasConcept C90805587 @default.
- W4288104915 hasConceptScore W4288104915C154945302 @default.
- W4288104915 hasConceptScore W4288104915C162324750 @default.
- W4288104915 hasConceptScore W4288104915C204321447 @default.
- W4288104915 hasConceptScore W4288104915C2524010 @default.
- W4288104915 hasConceptScore W4288104915C2776461190 @default.
- W4288104915 hasConceptScore W4288104915C33923547 @default.
- W4288104915 hasConceptScore W4288104915C38652104 @default.
- W4288104915 hasConceptScore W4288104915C41008148 @default.
- W4288104915 hasConceptScore W4288104915C41608201 @default.
- W4288104915 hasConceptScore W4288104915C48145219 @default.
- W4288104915 hasConceptScore W4288104915C50522688 @default.
- W4288104915 hasConceptScore W4288104915C53537400 @default.
- W4288104915 hasConceptScore W4288104915C547764534 @default.
- W4288104915 hasConceptScore W4288104915C80444323 @default.
- W4288104915 hasConceptScore W4288104915C90805587 @default.
- W4288104915 hasLocation W42881049151 @default.
- W4288104915 hasOpenAccess W4288104915 @default.
- W4288104915 hasPrimaryLocation W42881049151 @default.
- W4288104915 hasRelatedWork W1230873221 @default.
- W4288104915 hasRelatedWork W2560219888 @default.
- W4288104915 hasRelatedWork W2798009317 @default.
- W4288104915 hasRelatedWork W2920052239 @default.
- W4288104915 hasRelatedWork W2985392712 @default.
- W4288104915 hasRelatedWork W3036348210 @default.
- W4288104915 hasRelatedWork W3099354896 @default.
- W4288104915 hasRelatedWork W3133560342 @default.
- W4288104915 hasRelatedWork W4246455502 @default.
- W4288104915 hasRelatedWork W4287599800 @default.
- W4288104915 isParatext "false" @default.
- W4288104915 isRetracted "false" @default.
- W4288104915 workType "article" @default.