著者

上田 修一

出版者

慶應義塾大学

雑誌

Library and information science (ISSN:03734447)

巻号頁・発行日

vol.41, pp.17-25, 1999

被引用文献数

3

---

短報The word "kansei" originally means "sensitivity" in Japanese. A somewhat different andexpanded meaning of this word began to be used in Japan, at first in the area of marketing, in1984. Gradually it had been used in various fields, and eventually became a buzzword of the1980s. "Kansei" is an intentionally ambiguous word, meaning some ambiguous and illogicalway of accepting sensible objects, something like "feeling", "sensitivity", and "impression" inEnglish. In 1990s, the concept of "kansei" was introduced to information retrieval research. Experiments with "kansei" keywords consisting of adjectives, have been carried out for the retrievalof pictures and music. Normally steps of constructing and retrieving a "kansei" database are asfollows: (1) to assign "kansei" keywords to images manually; (2) to relate the "kansei" keywords thusassigned with the characteristics (ex. color) extracted from the images; (3) to create a databaseconsisting of images, their characteristics and "kansei" keywords; (4) to retrieve images with"kansei" keywords. Many experiments have shown that such a method of creating databases achieves theoptimal performance in small size databases. Thus, it will be most probable that it cannot beapplied to operational databases.