FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2767994764> ?p ?o ?g. }

• W2767994764 abstract "Conceptual Hierarchies Arise from the Dynamics of Learning and Processing: Insights from a Flat Attractor Network Christopher M. O’Connor (cmoconno@uwo.ca) Ken McRae (mcrae@uwo.ca) Department of Psychology, University of Western Ontario London, Ontario, Canada, N6A 6C2 George S. Cree (gcree@utsc.utoronto.ca) Department of Life Sciences, University of Toronto at Scarborough Toronto, Ontario, Canada, M1C 1A4 computed representations using a cosine similarity metric. The correlation between cosine and participants' rating of typicality was computed for each category. This correlation was high for most of the 20 superordinate categories, and predictions of typicality ratings were basically equivalent for the model and family resemblance. The structure of people's conceptual knowledge of concrete nouns has traditionally been viewed as hierarchical (Collins & Quillian, 1969). This has occurred primarily because human performance in many tasks (e.g., typicality rating and superordinate to exemplar priming) appear to implicate hierarchies. Thus hierarchies have been transported, rather transparently, into theories of the structure of knowledge. The goal of the present research is to show that behavior that, on the surface, appears to demand a hierarchical model can be simulated using a feature-based attractor network that contains a single level of semantics. In addition, this model provides insight into the role of similarity on the temporal dynamics of conceptual processing. Temporal Dynamics of Similarity A number of experiments have shown roughly equal superordinate to exemplar priming (fruit priming cherry) for high, medium, and low typicality exemplars. Paradoxically, other studies show that basic-level concepts must be highly similar to support priming (and attractor networks simulate these effects). We conducted an experiment in which we found that the magnitude of priming was indeed similar for high, medium and low typicality items. Priming simulations showed the same result. Unlike features of basic-level concepts, superordinate features are partially activated from a wordform, rather than being activated essentially to 1 or 0 (due to a superordinate's mapping from one wordform to many featural representations). Because the feature activations are on the active part of the sigmoidal function, it is easy for a network to move from a superordinate representation to the representation of one of its exemplars, resulting in equivalent priming effects regardless of typicality. Thus, the model's temporal dynamics provide novel insight into these seemingly inconsistent results. This research shows that a flat feature-based attractor network produces emergent behavior that accounts for human results that have previously been viewed as requiring a hierarchical representational structure. Network Architecture and Training The network mapped from 30 wordform nodes (representing spelling/sound of a word) to 2349 semantic feature nodes (there were no taxonomic features such as in the network; these were used to establish 20 superordinate categories). Each feature unit corresponded to a semantic feature (e.g., ) from McRae et al.’s (2005) norms. The feature nodes were fully inter-connected with no self-connections. Thus, no hierarchy was implemented transparently in the model. All 541 basic-level concepts were taken from the norms. Using continuous recurrent backpropagation, the network learned to map a 3-unit wordform for each basic-level concept to semantic features representing that concept. Superordinate concept learning was more complex because there were no unique target representations for them. On each superordinate learning trial, a wordform was paired with the representation of one of its exemplars. For example fruit was paired with the features of apple on one trial, cherry on another, etc. Crucially, exemplar representations were paired with the corresponding superordinate wordform equally often, so that typicality was not built into the training regime, and the network developed superordinate representations based on experience with the exemplars. Acknowledgments This research was supported by NSERC OGP0155704 and NIH DC0418 and MH6051701 grants to Ken McRae. References Collins, A. M., & Quillian, M. R. (1969). Retrieval time from semantic memory. Journal of Verbal Learning and Verbal Behavior, 8, 240-247. McRae, K., Cree, G. S., Seidenberg, M. S., & McNorgan, C. (2005). Semantic feature production norms for a large set of living and nonliving things. Behavior Research Methods, 37, 547-559. Typicality Any model of this sort must account for typicality ratings. This was simulated by computing the meaning of a superordinate (e.g., fruit), computing the meaning of an exemplar (e.g., cherry), then calculating the similarity of the" @default.
• W2767994764 created "2017-11-17" @default.
• W2767994764 creator A5002056928 @default.
• W2767994764 creator A5006501986 @default.
• W2767994764 creator A5085669226 @default.
• W2767994764 date "2006-01-01" @default.
• W2767994764 modified "2023-09-23" @default.
• W2767994764 title "Conceptual Hierarchies Arise from the Dynamics of Learning and Processing: Insights from a Flat Attractor Network" @default.
• W2767994764 cites W2003240077 @default.
• W2767994764 cites W2078894097 @default.
• W2767994764 hasPublicationYear "2006" @default.
• W2767994764 type Work @default.
• W2767994764 sameAs 2767994764 @default.
• W2767994764 citedByCount "2" @default.
• W2767994764 countsByYear W27679947642012 @default.
• W2767994764 crossrefType "journal-article" @default.
• W2767994764 hasAuthorship W2767994764A5002056928 @default.
• W2767994764 hasAuthorship W2767994764A5006501986 @default.
• W2767994764 hasAuthorship W2767994764A5085669226 @default.
• W2767994764 hasConcept C100701293 @default.
• W2767994764 hasConcept C103278499 @default.
• W2767994764 hasConcept C115961682 @default.
• W2767994764 hasConcept C119857082 @default.
• W2767994764 hasConcept C145912823 @default.
• W2767994764 hasConcept C154945302 @default.
• W2767994764 hasConcept C15744967 @default.
• W2767994764 hasConcept C162324750 @default.
• W2767994764 hasConcept C176217482 @default.
• W2767994764 hasConcept C180747234 @default.
• W2767994764 hasConcept C184337299 @default.
• W2767994764 hasConcept C188147891 @default.
• W2767994764 hasConcept C19417346 @default.
• W2767994764 hasConcept C199360897 @default.
• W2767994764 hasConcept C204321447 @default.
• W2767994764 hasConcept C21547014 @default.
• W2767994764 hasConcept C41008148 @default.
• W2767994764 hasConcept C48164120 @default.
• W2767994764 hasConcept C5274069 @default.
• W2767994764 hasConcept C59822182 @default.
• W2767994764 hasConcept C77805123 @default.
• W2767994764 hasConcept C78022038 @default.
• W2767994764 hasConcept C81444415 @default.
• W2767994764 hasConcept C86803240 @default.
• W2767994764 hasConceptScore W2767994764C100701293 @default.
• W2767994764 hasConceptScore W2767994764C103278499 @default.
• W2767994764 hasConceptScore W2767994764C115961682 @default.
• W2767994764 hasConceptScore W2767994764C119857082 @default.
• W2767994764 hasConceptScore W2767994764C145912823 @default.
• W2767994764 hasConceptScore W2767994764C154945302 @default.
• W2767994764 hasConceptScore W2767994764C15744967 @default.
• W2767994764 hasConceptScore W2767994764C162324750 @default.
• W2767994764 hasConceptScore W2767994764C176217482 @default.
• W2767994764 hasConceptScore W2767994764C180747234 @default.
• W2767994764 hasConceptScore W2767994764C184337299 @default.
• W2767994764 hasConceptScore W2767994764C188147891 @default.
• W2767994764 hasConceptScore W2767994764C19417346 @default.
• W2767994764 hasConceptScore W2767994764C199360897 @default.
• W2767994764 hasConceptScore W2767994764C204321447 @default.
• W2767994764 hasConceptScore W2767994764C21547014 @default.
• W2767994764 hasConceptScore W2767994764C41008148 @default.
• W2767994764 hasConceptScore W2767994764C48164120 @default.
• W2767994764 hasConceptScore W2767994764C5274069 @default.
• W2767994764 hasConceptScore W2767994764C59822182 @default.
• W2767994764 hasConceptScore W2767994764C77805123 @default.
• W2767994764 hasConceptScore W2767994764C78022038 @default.
• W2767994764 hasConceptScore W2767994764C81444415 @default.
• W2767994764 hasConceptScore W2767994764C86803240 @default.
• W2767994764 hasIssue "28" @default.
• W2767994764 hasLocation W27679947641 @default.
• W2767994764 hasOpenAccess W2767994764 @default.
• W2767994764 hasPrimaryLocation W27679947641 @default.
• W2767994764 hasRelatedWork W113694692 @default.
• W2767994764 hasRelatedWork W1534396039 @default.
• W2767994764 hasRelatedWork W1974275080 @default.
• W2767994764 hasRelatedWork W1989415743 @default.
• W2767994764 hasRelatedWork W2003663929 @default.
• W2767994764 hasRelatedWork W2009103014 @default.
• W2767994764 hasRelatedWork W2127757994 @default.
• W2767994764 hasRelatedWork W2162640456 @default.
• W2767994764 hasRelatedWork W2167981241 @default.
• W2767994764 hasRelatedWork W2502555949 @default.
• W2767994764 hasRelatedWork W2580972828 @default.
• W2767994764 hasRelatedWork W2581390987 @default.
• W2767994764 hasRelatedWork W2587370305 @default.
• W2767994764 hasRelatedWork W2619570484 @default.
• W2767994764 hasRelatedWork W2622634010 @default.
• W2767994764 hasRelatedWork W2625657005 @default.
• W2767994764 hasRelatedWork W2767468513 @default.
• W2767994764 hasRelatedWork W2958689161 @default.
• W2767994764 hasRelatedWork W43844756 @default.
• W2767994764 hasRelatedWork W3142766148 @default.
• W2767994764 hasVolume "28" @default.
• W2767994764 isParatext "false" @default.
• W2767994764 isRetracted "false" @default.
• W2767994764 magId "2767994764" @default.
• W2767994764 workType "article" @default.