Unsupervised Qualitative Scoring for Binary Item Features

• Published in: 2019 IEEE International Conference on Data Mining (ICDM) pp. 1114 - 1119
• Main Authors: Ichikawa, Koji, Tamano, Hiroshi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2019
• Subjects: collaborative filtering; label enhancement; topic model; unsupervised learning
• ISSN: 2374-8486
• DOI: 10.1109/ICDM.2019.00133

The qualitative score of tags is widely used to describe which product is better in terms of the given property. For example, in a restaurant-navigation site, properties such as food, location, and mood are given in the form of numerical values, representing the goodness of each aspect. In this paper, we propose a novel approach to estimate the qualitative score from the binary features of products. Based on a natural assumption that an item with a better property is more popular among users who prefer that property, in short, "experts know best", we introduce one discriminative and two generative models with which user preferences and item-qualitative scores are inferred from user-item interactions. Our approach contributes to resolving the following difficulties: (1) no supervised data for the score estimation, (2) implicit user purpose, and (3) irrelevant tag contamination. We evaluate our models by using two artificial datasets and two real-world datasets of movie and book ratings and observe that our models outperform baseline models.