2017 年 137 巻 7 号 p. 974-979
There are many online shops in the Internet and a lot of persons buy products in those shops. Many online shops provide advertisements and item recommendations to attract their customers, but to achieve satisfactory customer experience, shops have to change advertisement and recommendation strategies for every customer's intents, since his/her intent decides whether provided advertisements and recommendations are useful or not. Hence it is important for online shops to estimate customer's intents. One of the resources we can use to estimate is an access log, including information on customer's movement in the online shop. If the movements depend on customer's intents, we can extract knowledge on them from the access logs. However, tracking user's changeable intents is difficult for existing time-series topic models since there would be few user activities in a time slice. In this paper, we propose a new topic model combining a topic model for behavior analysis with a model for short text analysis. In experiments, we show that our new model can track user's intents appropriately.
J-STAGEがリニューアルされました! https://www.jstage.jst.go.jp/browse/-char/ja/