作者: Douglas Hayhoe
DOI: 10.1016/S0020-7373(05)80069-0
关键词: Artificial intelligence 、 Pairwise comparison 、 Computer science 、 Natural language processing 、 Goodness of fit 、 Subject (documents) 、 Data mining 、 Selection (linguistics) 、 Categorization 、 Sorting
摘要: Several researchers have conducted sorting experiments or pairwise comparisons with a database of menu items in order to form coherent categories. However, these all contained one more the following potential weaknesses: (1) they used only particular database; (2) too few subjects; (3) uncritically scaling technique clusters; (4) did not conduct an experimental comparison categories formed. In present research, were 48 subjects and four 48-item databases: clothes, furniture, occupations, sports. Latent partition analysis hierarchical clustering (Ward's method group average linkage) These placed into “pull-down” system two conditions; titles chosen by each individual subject; investigator. Two other conditions added: formed software design experts; subject for his her own work. within-subjects performed. The sorting-based investigator superior expert selection times, errors, ”goodness fit” ratings, memory recall errors. A detailed showed that “miscategorization” errors vague category than categories, while both overlapping