Dialogue-based Cyber Salesperson for Music Recommendation

초록

Recently many customers move to online shopping malls, however, they are not satisfied with the service of malls. One of the reasons is that every shopping mall has a different way to operate, most customers have to adjust themselves for the most of shopping malls in order to achieve their goals. What is worse, beginners have had many troubles with their choice and options in purchasing information on online shopping malls. For these reasons web-based dialogue systems using natural language to help their customers have been appeared to introduce product information and website. It is, however, still difficult for customers to make a decision. Therefore, it is necessary that the system can talk with customers to gather information about them by having a conversation with people. This paper proposes a recommender system which can interact with shoppers. By having a conversation with shoppers, we can automatically extract user’s demographic properties such as age, sex, citizenship, job, and hobby to predict user’s preference on music. To accomplish this extraction, we have used the user classification based on Bayes’s rule. In addition, we extract the preference on music data for a particular user who is not being included in the classification of user models. For experimentations, this system has been applied with Bayes’s rule for user modeling. We have chosen the music CD shopping mall as an example to prove the efficiency of this system. We have used almost 5200 members of preference data on music for training our system