A Search Method That Can Dynamically Learn User´s Subjective Feeling

As an approach to search /retrieve such objects as pictures, music, perfumes and apparels on the Internet, sensitivity-vectors or kansei-vectors are useful since textual keywords are not sufficient to find objects that users want. The sensitivity-vector is an array of values each indicates a degree of feeling or impression represented as sensitivity word or kansei word. However, due to the gap between user´s subjective sensitivity (image, feeling) value and the corresponding one stored in the database, even such an approach is not enough to search what users want. This paper proposes a search method to automatically and dynamically decrease such gaps by estimating a subjective criterion deviation (we call “SCD”) using the user´s search history and fuzzy modeling. The proposed method can avoid users´ burden caused by conventional methods such as requiring questionnaires. This method can also reflect the dynamic change of user´s preference which cannot be accomplished by using questionnaires. For the evaluation, an experiment was done building and using a perfume search system. Through observing the transition of the deviation elimination degree, it was clarified that the proposed method is effective. In the experiment, the machine could learn users´ subjective criteria deviation as well as its dynamic change caused by factors such as user´s preference, if the learning rate is well adjusted.