Searching for the perfect match: a comparison of free sorting results for images by human subjects and by latent semantic analysis techniques

Author

Börner, Katy

Author_Institution

Sch. of Libr. & Inf. Sci., Indiana Univ., Bloomington, IN, USA

fYear

2000

fDate

2000

Firstpage

192

Lastpage

197

Abstract

Information visualization can improve the access and manipulation of digital data if it conforms to human expectations, and supports the particular interface, task, data, and user. Detailed usability studies are required to prove this assumption. The paper reports a usability study of the data analysis algorithm applied in the LVis, Digital Library Visualizer that spatially visualizes search results derived from user queries of digital library collections. In particular, it presents a comparison of free sorting results for image data done by 20 human subjects and the data analysis result derived via Latent Semantic Analysis from the textual image descriptions. The results of the study suggest that people, if asked to categorize with no constraints, will use features that are often derivable from textual metadata. Human clustering strategies and resulting demands for image browsing systems are discussed and future work is outlined

Keywords

data visualisation; digital libraries; human factors; information retrieval; sorting; user interfaces; visual databases; Digital Library Visualizer; LVis; Latent Semantic Analysis; data analysis algorithm; digital data access; digital library collections; free sorting results; future work; human clustering strategies; human expectations; human subjects; image browsing systems; image data; information visualization; latent semantic analysis techniques; perfect match; spatial visualization; textual image descriptions; textual metadata; usability studies; user queries; Data visualization; Humans; Image analysis; Image retrieval; Information analysis; Software libraries; Sorting; Support vector machine classification; Support vector machines; Usability;

fLanguage

English

Publisher

ieee

Conference_Titel

Information Visualization, 2000. Proceedings. IEEE International Conference on

Conference_Location

London

ISSN

1093-9547

Print_ISBN

0-7695-0743-3

Type

conf

DOI

10.1109/IV.2000.859755

Filename

859755