Comparison Between Semantic Similarities of Inconsistent Objects in Gist-Explicit and Gist-Inexplicit Scenes During Eye-Movements

DOI: 10.5176/2251-1865_CBP14.45

Authors: KangWoo Lee, Myoung-Hee Shin, Jeong-soo Lee, and Hyunseung Choo

Abstract:

The role of semantic consistency is controversial in attentional prioritization. We examine attentional prioritization by comparing three different similarity measures – gist-to-object, whole objet-to-object and local object-to-object similarities that are constructed from Latent Semantic Analysis (LSA). The comparison was done in gist-explicit image and gist-inexplicit image conditions. We found that eye fixation is guided by consistency of scene gist (context) when scene gist is clearly shown. No inconsistency effect on attentional prioritization was found. The results imply that the gist-to-object relation would be more swiftly extracted and influential for scene understanding than object-to-object relation.

Keywords: component; attentional prioritization; semantic similarity; Latent Semantic Analysis; eye tracking; fixation order