FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W10816> ?p ?o ?g. }

• W10816 abstract "Evaluating the Contribution of Intra-Linguistic and Extra-Linguistic Data to the Structure of Human Semantic Representations Mark Andrews (m.andrews@ucl.ac.uk) Gabriella Vigliocco (g.vigliocco@ucl.ac.uk) David Vinson (d.vinson@ucl.ac.uk) Department of Psychology, University College London, 26 Bedford Way London, WC1H 0AP United Kingdom any one of these data types had been considered inde- pendently and to the exclusion of the other. To address this concern, we considered the combined effects of both sources of data and introduced a probabilistic model that learns semantic representations on the basis of both at- tributional and distributional data simultaneously. We then compared this model with probabilistic models that learn semantic representations from each data source in- dependently. In our above mentioned work, we did not provide an analysis of how well the semantic representations learned by our model predict human data. The primary aim of this paper is to address this issue. For this purpose, we have also found it necessary to elaborate and extend upon the models that we previously used. As such, in what follows, we provide Bayesian models of semantic representations that are learned from either attributional data or distributional data, or from both in combina- tion. We then evaluate the validity of these models us- ing three human-based measures of semantic similarity: word-association norms, semantic-priming results from a lexical decision task, and interference patterns from a picture-word interference task. Abstract We describe Bayesian models that learn semantic rep- resentations from either extra-linguistic data or intra- linguistic data, or from both in combination. We evalu- ate the validity of these models using three human-based measures of semantic similarity. The results provide strong evidence for the hypothesis that human semantic representations are the product of the statistical combi- nation of extra- and intra-linguistic sources of data. Introduction For the purposes of this paper, we use the term seman- tic representation to refer to a language user’s mental or cognitive representation of the meaning of words. We in- formally define this as the knowledge that allows the lan- guage user to infer, amongst other things, which words are similar or identical in meaning, what are the semantic or ontological categories to which a word belongs, what (if anything) are the referents of a word. Our general aim is to consider how both extra-linguistic and intra- linguistic data can be used to acquire this knowledge. Extra-linguistic, or attributional data, is data that is de- rived from our perception and interaction with the phys- ical world, and in particular, from the perceived physi- cal attributes or properties associated with the referents of words 1 . In contrast, intra-linguistic, or distributional data, is derived from the statistical characteristics within a language itself, or how a given word is distributed across different spoken or written texts 2 In previous literature, it has been repeatedly demon- strated that semantic representations can be learned from either attributional data alone, e.g. McRae, Sa, and Seidenberg (1997); Vigliocco, Vinson, Lewis, and Garrett (2004); McClelland and Rogers (2003), or dis- tributional data alone, e.g. Lund and Burgess (1996); Landauer and Dumais (1997); Griffiths and Steyvers (2002). However, in previous work of our own (Andrews, Vigliocco, & Vinson, 2005), we described how, for the most part throughout this literature, the contribution of Model Description We provide Bayesian models that learn semantic repre- sentations from examples of attributional data, or from distributional data, or from both combined. The proba- bilistic models we employ for each of the various data types are described graphically in Figure 1. The at- tributional model (leftmost) describes any given word w f as a probability distribution over a set of binary attributes, such that {y m [f ] : 1 ≤ m ≤ M [f ] } is a set of bit vectors, each being an instance of the referent of the word w f . These probability distributions are compositions of a basic repertoire of latent distributions ψ = {ψ 1 . . . ψ K Att } that intuitively correspond to clus- ters of interrelated attributes each describing basic char- acteristics of the attributional data. The distributional model (second left) describes texts as multinomial dis- tribution over words, such that {w n [t] : 1 ≤ n ≤ N [t] } is a sample of words from text t. These distributions are compositions of latent distributions φ = {φ 1 . . . φ K Dist } that intuitively correspond to discourse-topics in a cor- pus of text. The combined model (second right) de- scribes texts as probability distributions over words, and words as distributions over attributes. These distribu- For example, the word apple refers to objects in the world whose perceived attributes or properties include being red or green, round, shiny, smooth, crunchy, juicy, sweet, tasty, etc We use the term text here in a very general sense to refer to any coherent and self-contained piece of written or spo- ken language. This could include, for example, a newspaper article, a spoken conversation, a letter or email message, an essay, a speech, etc." @default.
• W10816 created "2016-06-24" @default.
• W10816 creator A5012340808 @default.
• W10816 creator A5024406485 @default.
• W10816 creator A5063093565 @default.
• W10816 date "2007-01-01" @default.
• W10816 modified "2023-09-23" @default.
• W10816 title "Evaluating the Contribution of Intra-Linguistic and Extra-Linguistic Data to the Structure of Human Semantic Representations" @default.
• W10816 cites W125892352 @default.
• W10816 cites W1975849605 @default.
• W10816 cites W1981617416 @default.
• W10816 cites W1983578042 @default.
• W10816 cites W1993812374 @default.
• W10816 cites W2113881303 @default.
• W10816 cites W2126163471 @default.
• W10816 cites W2590882584 @default.
• W10816 hasPublicationYear "2007" @default.
• W10816 type Work @default.
• W10816 sameAs 10816 @default.
• W10816 citedByCount "2" @default.
• W10816 countsByYear W108162013 @default.
• W10816 countsByYear W108162014 @default.
• W10816 crossrefType "proceedings-article" @default.
• W10816 hasAuthorship W10816A5012340808 @default.
• W10816 hasAuthorship W10816A5024406485 @default.
• W10816 hasAuthorship W10816A5063093565 @default.
• W10816 hasConcept C126398093 @default.
• W10816 hasConcept C130318100 @default.
• W10816 hasConcept C138885662 @default.
• W10816 hasConcept C154945302 @default.
• W10816 hasConcept C15744967 @default.
• W10816 hasConcept C162324750 @default.
• W10816 hasConcept C169760540 @default.
• W10816 hasConcept C169900460 @default.
• W10816 hasConcept C187736073 @default.
• W10816 hasConcept C204321447 @default.
• W10816 hasConcept C2780451532 @default.
• W10816 hasConcept C41008148 @default.
• W10816 hasConcept C41895202 @default.
• W10816 hasConcept C49937458 @default.
• W10816 hasConcept C90805587 @default.
• W10816 hasConceptScore W10816C126398093 @default.
• W10816 hasConceptScore W10816C130318100 @default.
• W10816 hasConceptScore W10816C138885662 @default.
• W10816 hasConceptScore W10816C154945302 @default.
• W10816 hasConceptScore W10816C15744967 @default.
• W10816 hasConceptScore W10816C162324750 @default.
• W10816 hasConceptScore W10816C169760540 @default.
• W10816 hasConceptScore W10816C169900460 @default.
• W10816 hasConceptScore W10816C187736073 @default.
• W10816 hasConceptScore W10816C204321447 @default.
• W10816 hasConceptScore W10816C2780451532 @default.
• W10816 hasConceptScore W10816C41008148 @default.
• W10816 hasConceptScore W10816C41895202 @default.
• W10816 hasConceptScore W10816C49937458 @default.
• W10816 hasConceptScore W10816C90805587 @default.
• W10816 hasIssue "29" @default.
• W10816 hasLocation W108161 @default.
• W10816 hasOpenAccess W10816 @default.
• W10816 hasPrimaryLocation W108161 @default.
• W10816 hasRelatedWork W125892352 @default.
• W10816 hasRelatedWork W177177099 @default.
• W10816 hasRelatedWork W2072323436 @default.
• W10816 hasRelatedWork W2171086879 @default.
• W10816 hasRelatedWork W2251041941 @default.
• W10816 hasRelatedWork W2398161125 @default.
• W10816 hasRelatedWork W2551572025 @default.
• W10816 hasRelatedWork W2575124246 @default.
• W10816 hasRelatedWork W2596535248 @default.
• W10816 hasRelatedWork W2599992093 @default.
• W10816 hasRelatedWork W2767997808 @default.
• W10816 hasRelatedWork W2913142708 @default.
• W10816 hasRelatedWork W2913316250 @default.
• W10816 hasRelatedWork W2944617758 @default.
• W10816 hasRelatedWork W2950018201 @default.
• W10816 hasRelatedWork W2954758982 @default.
• W10816 hasRelatedWork W2994896873 @default.
• W10816 hasRelatedWork W3102628894 @default.
• W10816 hasRelatedWork W3172797824 @default.
• W10816 hasRelatedWork W75561580 @default.
• W10816 hasVolume "29" @default.
• W10816 isParatext "false" @default.
• W10816 isRetracted "false" @default.
• W10816 magId "10816" @default.
• W10816 workType "article" @default.