Event Title
Data Mining a Hypernetwork of Features of Early-learned Nouns to Find Nested and Overlapping Categories of Early Nouns: Comparison with the Binary Network and the Clique Percolation Method
Presentation Type
Poster/Portfolio
Presenter Major(s)
Psychology, Communications
Mentor Information
Josita Maouene
Department
Psychology
Location
Henry Hall Atrium 61
Start Date
10-4-2013 11:00 AM
End Date
10-4-2013 12:00 PM
Keywords
Social Science
Abstract
Former work in the development of semantic networks described the hierarchical information contained in the overlapping features of a binary network of 132 nouns using a method, the clique percolation, to interpret the global organization and the local structures of the network. However, both are limited as they can only yield minimal semantic information such as the degree of connectivity of local structures, whether a link of similarity exist or not between two neighbors or identify cliques with unspecified semantics. To unveil the richness of the semantic organization of this data and the relationships of overlap and containment latent in those structures, and thus to answer categorical questions like what are the semantic features that connect zebra, horse and goat in the animal-like cluster, or what is a more general feature, eats_grass or has_a_tail, we propose to use a different type of representation: hypernetwork and a different formalism: Formal Concept Analysis.
Data Mining a Hypernetwork of Features of Early-learned Nouns to Find Nested and Overlapping Categories of Early Nouns: Comparison with the Binary Network and the Clique Percolation Method
Henry Hall Atrium 61
Former work in the development of semantic networks described the hierarchical information contained in the overlapping features of a binary network of 132 nouns using a method, the clique percolation, to interpret the global organization and the local structures of the network. However, both are limited as they can only yield minimal semantic information such as the degree of connectivity of local structures, whether a link of similarity exist or not between two neighbors or identify cliques with unspecified semantics. To unveil the richness of the semantic organization of this data and the relationships of overlap and containment latent in those structures, and thus to answer categorical questions like what are the semantic features that connect zebra, horse and goat in the animal-like cluster, or what is a more general feature, eats_grass or has_a_tail, we propose to use a different type of representation: hypernetwork and a different formalism: Formal Concept Analysis.