行動計量学
Online ISSN : 1880-4705
Print ISSN : 0385-5481
ISSN-L : 0385-5481
原著
漢字記憶検索過程を支える類似性空間の計算モデル
―IsomapとMDSの比較―
地村 弘二楠見 孝栗山 直子松岡 隆史中川 正宣
著者情報
ジャーナル フリー

2003 年 30 巻 2 号 p. 199-212

詳細
抄録
The aim of the present study was to investigate the role of similarity in memory retrieval processes by employing similarity-judgment tasks and computational models. Five conditions of similarity-judgment tasks of Kanji characters were first conducted to 195 university students. Correlational analyses indicated that the similarity among the targets was primarily dependent on their figural feature rather than the semantic or phonetic feature, and moreover, the similarity was significantly correlated with the retrieval order of the targets during a memory retrieval task (Jimura et al., 1999). Then, the similarity space constituted by the shape of the targets was applied to two computational models in subsymbolic level : the multi-dimensional scaling (MDS) and the Isometric feature mapping (Isomap). Dimensional reduction of the shape-based similarity space indicated that the Isomap more accurately represented the similarity derived from the judgment-task and the retrieval pattern than the MDS. These results suggest the importance of the similarity in memory retrieval processes and the superiority of Isomap as a subsymbolic model of similarity.
著者関連情報
© 2003 日本行動計量学会
前の記事 次の記事
feedback
Top