Matches in SemOpenAlex for { <https://semopenalex.org/work/W425330175> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W425330175 abstract "The resolution of ambiguity is one of the most difficult problems in natural language analysis. This thesis describes an approach to the problem of ambiguity resolution that is based on the loglinear model. The advantage of using a loglinear model is that it takes into account the effects of combinations of feature values, as well as the main effects of individual feature values. Experiments concerning three types of ambiguity are described: modeling unknown words, lexical categorial ambiguity, and prepositional phrase attachment ambiguity.The first series of experiments show that the statistical model of unknown words obtains a higher accuracy than the method described by (Weischedel et al, 1993), which assumes independence between the different features.The second series of experiments on lexical categorial ambiguity identify unknown words as an important source of error in Part-of-Speech tagging. The loglinear method raises the median accuracy significantly from a baseline method that uses the prior distribution of infrequent words. A simpler model of unknown words that assumes independence between the different features brings about significantly less improvement. An additional experiment on Part-of-Speech tagging with the loglinear model is described.A final set of experiments concerns Prepositional Phrase (PP) attachment ambiguity. On the Verb NP PP pattern, the loglinear method does not perform significantly better than the strategy of lexical association presented by (Hindle and Rooth, 1993). The Verb NP NP PP pattern with three possible attachment sites, including two sites that have the same syntactic category, on the other hand, shows a marked improvement for the loglinear method.These results show that the loglinear method results in higher accuracy than simpler methods which rely on a single feature, or which assume independence between multiple features. Still, a gap remains between the accuracy of the loglinear method, and human performance on ambiguity resolution tasks. This suggests that in order to achieve very high accuracy, it is necessary to consider higher-order factors, and to integrate statistical natural language processing methods with manually acquired human expertise." @default.
- W425330175 created "2016-06-24" @default.
- W425330175 creator A5030184120 @default.
- W425330175 creator A5037897140 @default.
- W425330175 date "1995-01-01" @default.
- W425330175 modified "2023-09-27" @default.
- W425330175 title "A statistical approach to syntactic ambiguity resolution" @default.
- W425330175 hasPublicationYear "1995" @default.
- W425330175 type Work @default.
- W425330175 sameAs 425330175 @default.
- W425330175 citedByCount "0" @default.
- W425330175 crossrefType "journal-article" @default.
- W425330175 hasAuthorship W425330175A5030184120 @default.
- W425330175 hasAuthorship W425330175A5037897140 @default.
- W425330175 hasConcept C105795698 @default.
- W425330175 hasConcept C138885662 @default.
- W425330175 hasConcept C14279187 @default.
- W425330175 hasConcept C154945302 @default.
- W425330175 hasConcept C163175372 @default.
- W425330175 hasConcept C199360897 @default.
- W425330175 hasConcept C204321447 @default.
- W425330175 hasConcept C2776224158 @default.
- W425330175 hasConcept C2776397901 @default.
- W425330175 hasConcept C2776401178 @default.
- W425330175 hasConcept C2777559092 @default.
- W425330175 hasConcept C2780522230 @default.
- W425330175 hasConcept C33923547 @default.
- W425330175 hasConcept C35651441 @default.
- W425330175 hasConcept C41008148 @default.
- W425330175 hasConcept C41895202 @default.
- W425330175 hasConcept C60229501 @default.
- W425330175 hasConcept C70519679 @default.
- W425330175 hasConcept C76155785 @default.
- W425330175 hasConceptScore W425330175C105795698 @default.
- W425330175 hasConceptScore W425330175C138885662 @default.
- W425330175 hasConceptScore W425330175C14279187 @default.
- W425330175 hasConceptScore W425330175C154945302 @default.
- W425330175 hasConceptScore W425330175C163175372 @default.
- W425330175 hasConceptScore W425330175C199360897 @default.
- W425330175 hasConceptScore W425330175C204321447 @default.
- W425330175 hasConceptScore W425330175C2776224158 @default.
- W425330175 hasConceptScore W425330175C2776397901 @default.
- W425330175 hasConceptScore W425330175C2776401178 @default.
- W425330175 hasConceptScore W425330175C2777559092 @default.
- W425330175 hasConceptScore W425330175C2780522230 @default.
- W425330175 hasConceptScore W425330175C33923547 @default.
- W425330175 hasConceptScore W425330175C35651441 @default.
- W425330175 hasConceptScore W425330175C41008148 @default.
- W425330175 hasConceptScore W425330175C41895202 @default.
- W425330175 hasConceptScore W425330175C60229501 @default.
- W425330175 hasConceptScore W425330175C70519679 @default.
- W425330175 hasConceptScore W425330175C76155785 @default.
- W425330175 hasLocation W4253301751 @default.
- W425330175 hasOpenAccess W425330175 @default.
- W425330175 hasPrimaryLocation W4253301751 @default.
- W425330175 hasRelatedWork W1525582138 @default.
- W425330175 hasRelatedWork W1748923414 @default.
- W425330175 hasRelatedWork W1856198964 @default.
- W425330175 hasRelatedWork W1930328034 @default.
- W425330175 hasRelatedWork W1965949675 @default.
- W425330175 hasRelatedWork W2016488422 @default.
- W425330175 hasRelatedWork W2024963007 @default.
- W425330175 hasRelatedWork W2060356208 @default.
- W425330175 hasRelatedWork W2070523965 @default.
- W425330175 hasRelatedWork W2071490478 @default.
- W425330175 hasRelatedWork W2158873310 @default.
- W425330175 hasRelatedWork W2186083909 @default.
- W425330175 hasRelatedWork W2250367932 @default.
- W425330175 hasRelatedWork W2316951331 @default.
- W425330175 hasRelatedWork W2398584126 @default.
- W425330175 hasRelatedWork W2952343510 @default.
- W425330175 hasRelatedWork W52022909 @default.
- W425330175 hasRelatedWork W98990824 @default.
- W425330175 hasRelatedWork W2107029693 @default.
- W425330175 hasRelatedWork W2398736350 @default.
- W425330175 isParatext "false" @default.
- W425330175 isRetracted "false" @default.
- W425330175 magId "425330175" @default.
- W425330175 workType "article" @default.