Profile Generation for TV Program Recommendation Based on Utterance Analysis
Yasufumi Takama, Yuki Muto · Journal of Advanced Computational Intelligence and Intelligent Informatics · 2009
This paper proposes a method for generating user profile that is to be used for TV program recommendation. The proposed method does not estimate user's interest in a TV program only based on its watching time as most of existing methods do, but also based on user's utterances by applying sentiment analysis. Three kinds of scores are calculated for each watched program, based on which fuzzy inference is performed to estimate its rating. After the estimation, profile structure is obtained by generating category and subcategory layers. This paper mainly focuses on estimation of users' ratings on TV programs. Experiments are performed with test subjects, and the results show the proposed method can improve the estimation accuracy of ratings compared with existing approach using only watching time. Although the proposed method includes language-dependent processing, it is expected the core of estimation rating can be applied to other languages.