基于 GRU 算法的音乐和词语的情感语义匹配算法-《计算机技术与发展》

文章信息/Info

作者:: 朱星浩¹ ; 胥备²; 1. 南京邮电大学计算机、软件、网络空间安全学院,江苏南京 210023;
2. 江苏省大数据安全与智能处理重点实验室(南京邮电大学),江苏南京 210023

Author(s):: ZHU Xing-hao¹ ; XU Bei²; 1. School of Computer Science and Technology,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;
2. Jiangsu Key Laboratory of Big Data Security and Intelligent Processing ( Nanjing University of Posts and Telecommunications) , Nanjing 210023,China

Keywords:: emotion; symbol; semantic matching; correlation degree; natural language processing

摘要:: 自然语言和音乐是人们表达情感和描绘事物的两种语义符号系统。分析和建立语言与音乐的语义关联不仅有助于提供更精确的文本和音乐的检索和推荐服务,还可以帮助研究者进一步理解情感语义。已有的研究主要关注自然语言和音乐的表层符号特征,较少考虑其语义含义,从而限制了基于自然语言和音乐语义关联的应用的精确性和可解释性。另一方面,部分应用,例如音乐情感的多标签分类,需要更精确的自然语言和音乐的语义关联。所以,分析和建立自然语言和音乐的语义关联对于面向文本和音乐的应用有较大的促进作用。文中提出了自然语言和音乐的情感语义关联度计算算法——DeepTransition( DT-GRU) 。该算法以词语和音乐片段分别作为自然语言和音乐的基本单元,以基本情感为音乐和词语的共同语义进行关联度计算。多组实验证明,相比于同类算法,DT-GRU 可以更合理地计算音乐和词语的情感关联度。

Abstract:: Natural language and music are two semantic symbol systems for people to express emotions and describe things. The analysis and establishment of semantic association between language and music not only helps to provide more accurate retrieval and recommendation services for text and music,but also helps researchers to further understand emotional semantics. The existing studies mainly focus on the surface symbolic features of natural language and music,and less consider their semantic meaning,which limits the accuracy and interpretability of the application based on the semantic association of natural language and music. On the other hand,some applications,such as multi-label classification of music emotion,need deeper semantic association between natural language and music.Therefore,the analysis and calculation of the semantic association between natural language and music can promote the application of text and music. In this paper,we propose a new algorithm——Deep Transition( DT-GRU) to calculate the emotional semantic association between natural language and music. In this algorithm, words and music segments are regarded as the basic units of natural language and music,and the basic emotion is the common semantic of music and words to calculate the association degree. A number of experiments show that compared with similar algorithms,DT-GRU can calculate the emotional correlation between music and words more rationally.