Research

Demands for online shopping sites are increasing because more products can be presented than in physical stores, and it provides vendors with more sales opportunities and more choices for consumers.

Many of these shopping sites, such as Amazon, adopt product recommendations such as collaborative filtering to lead users to products purchased by other users with similar preferences and encourage their purchase. However, a user's preference is possibly changed during shopping.

Therefore, a new product recommendation method considering preference dynamics is proposed in this study. In the proposed method, the factor which change a user's preference is acquired and utilized to extend time spent of the shopping site.

The proposed method applies clustering to the products that are selected by a user and figures out the trends of preference. The transition of user's preference is detected by the difference between the past clustering result and current one.

(Click this image to open PDF)

Consumer Generated Media(CGM) is growing rapidly and the amount of content is increasing. However, it is often difficult for users to extract important contents and the existence of contents recording their experiences can easily be forgotten.

As there are no methods or systems to indicate the subjective value of the contents or ways to reuse them, subjective annotation appending subjectivity, such as feelings and intentions, to contents is needed. Representation of subjectivity depends on not only verbal expression, but also nonverbal expression. Linguistically expressed annotation, typified by collaborative tagging in social bookmarking systems, has come into widespread use, but there is no system of nonverbally expressed annotation on the web.

We proposed the utilization of controllable avatars as a means of nonverbal expression of subjectivity, and confirmed the consistency of feelings elicited by avatars over time for an individual and in a group.