An approach to category classification of cosmetics reviews based on brand names

In this paper, we have investigated the relation among shopping items, especially the same shopping items of other brands, in consumers´ reviews, aiming to use the result for proposing a better recommendation method. We firstly analyzed the reviews of Japan Yahoo! shopping to investigate which shopping categories the relation among brands often appear in, and found that it often appears in the reviews of cosmetics, then, we set the target reviews as cosmetic revies and extracted the brand´s relation described in them. To extract the relation, we employed rule based knowledge and support vector machine. Through experiments, we have found the features working well to extract the relation.