In recent years, research has been tried in various ways to incorporate sensitivity (emotion, feeling) in music content service. At home and abroad in a number of commercial music services, concept of user sensitivity has been incorporated in order to make their users feel more friendly and comfortable. However, the number of emotion words applied to those services was too small to express the varied and complex emotions user wants. In addition, the number of emotional contents was not sufficient. In this paper, we proposed the music sensitivity model and evaluation method, as a key component of customized service which provides personalized content search and social recommendation features. The proposed model provides a number of emotional words, built based on a musical vocabulary freely expressed on the Web, and intuitive classification of emotional words including mood (human feeling), emotion (content feeling), situation, and property categories. The proposed evaluation method provides an intuitive user interface for users to easily express their sensitivity, what they feels after listening music contents. The experiment was performed on 20 subjects, tested for lexical classification and sensitivity evaluation tasks, and the results showed the proposed method are meaningfully close to user intuition.