基于双向情感分析的实时性音乐推荐系统设计

目前的音乐推荐系统,一般采用基于个体兴趣的推荐方法,这种方式虽然能满足大部分情景下的用户需求,但无法感知到用户实时性的心情变化。考虑到不同情绪状态下用户对于音乐的需求往往也会发生改变,提出一种基于双向情感分析的算法并构建了实际系统,实时分析用户的情感需求来进行音乐推荐。一方面基于音乐在频域的梅尔倒谱系数构建特征分类器完成歌曲的情感分类;另一方面通过获取用户在社交网络中的实时文本信息,基于自然语言理解分析出用户当前的情感需求,最终为用户产生音乐推荐列表,实现基于情境感知的实时音乐推荐。实验表明,使用该个性化推荐算法具有更高的准确性,用户群体可以获得更为满意的用户体验。

Designing of Real-time Music Recommendation SystemBased on Two-Way Emotion Analysis

The current music recommendation system generally adopts the recommendation method on the basis of individual interest. Such method can meet users’ needs under most situations, but it fails to perceive the real-time mood changes of users. In view of the users’ demands for music under different emotional state always changing, the paper puts forward a two-way emotion analysis algorithm and constructs a real system to recommend music through real-time analysis of users’ emotion needs. On the one hand, based on MEL cepstrum coefficient of the music in the frequency domain, a feature classifier is built to complete the music emotion classification. On the other hand, by capturing the users’ real-time text social information in social networks, the emotional needs of the users are analyzed all the time based on natural language understanding. Finally, a music recommendation list is produced to the user for a specific situation, achieving the real-time music recommendation based on situational awareness. The experiment results show that the user groups who adopt such personalized recommendation algorithm can obtain more satisfying user experience.