抄録
An apparel product search system was proposed, in which customers can refer to products through several stores. With the system, the user can search for apparel products by inputting category information and/or Kansei measure values. A unified format of category information of apparel products suited for the search was developed. The Kansei search was performed by using some measure values of Kansei words. The distance between the customer input of Kansei information and store side evaluation are calculated, then the closest products are shown. Semantic differential (SD) method and statistics analyses were used for selection of measure terms and evaluation of efficiency of the search. Generally, the Kansei evaluation of a product of the store side doesn't agree with customer evaluation. Some learning methods were investigated to improve the effectiveness of the search. In the learning, Kansei evaluations of individual customer and store side were made with some products beforehand. Statistical analysis was made with the results of the SD test to find effective methods. If learning is enough, the efficiency of search improved by using the linear multiple regression equation between store side and individual customer's evaluations.
