第15卷第1期 智能系统学报 Vol.15 No.1 2020年1月 CAAI Transactions on Intelligent Systems Jan.2020 D0L:10.11992tis.202001006 基于序列模型的音乐词曲匹配度智能评估算法 陈壮豪'，张茂清，郭为安2，康琦，汪镭 (1.同济大学电子与信息工程学院，上海201804：2.同济大学中德工程学院，上海201804) 摘要：情感匹配模型是一种常用于评价词曲匹配程度的方法；然而，单纯地依靠情感匹配模型无法对评价词 曲匹配度进行准确的评价。为改善此问题，提出了基于序列模型的词曲匹配度智能评估算法，其综合考虑词曲 情感和词曲间节奏关系以给出一个更加准确的词曲评估方法。基于公开词曲同步数据集，构建了音乐情感和 节奏正反例模型，并基于此模型将音乐切分成片段：进一步，将歌词和旋律片段分别通过歌词编码器和旋律编 码器进行编码，并将编码后具有上下语境的歌词特征和旋律特征输人词曲匹配解码器，解析词曲间特征关系， 判断词曲片段匹配程度。仿真结果表明：基于序列模型的词曲匹配度智能评估算法，相对于单纯的情感匹配模 型，能够更精确地评价词曲匹配程度，验证了本文提出算法的有效性。 关键词：音乐词曲：情感：节奏：序列模型：歌词编码器：旋律解码器：词曲匹配解码器：词曲匹配度： 中图分类号：TP393.04文献标志码：A文章编号：1673-4785(2020)01-0067-07 中文引用格式：陈壮豪，张茂清，郭为安，等.基于序列模型的音乐词曲匹配度智能评估算法.智能系统学报，2020,15(1)： 67-73. 英文引用格式：CHEN Zhuanghao,ZHANG Maoqing,GUO Weian,.etal.Music lyrics-melody intelligent evaluation algorithm based on sequence model Jl.CAAI transactions on intelligent systems,2020,15(1):67-73. Music lyrics-melody intelligent evaluation algorithm based on sequence model CHEN Zhuanghao',ZHANG Maoqing',GUO Weian',KANG Qi',WANG Lei (1.College of Electronics and Information Engineering,Tongji University,Shanghai 201804,China;2.Sino-German College of Ap- plied Sciences,Tongji University,Shanghai 201804,China) Abstract:Emotional matching model is a method often used to evaluate the degree of lyrics and melody matching. However,it cannot be accurately evaluated based on the emotion matching model.In order to improve it,this paper pro- poses an intelligent evaluation algorithm of lyrics-melody matching based on a sequence model,which comprehens- ively considers the emotion and the rhythm relationship between lyrics and melody to give an evaluation method for more accurate evaluation.Firstly,this paper researches and builds music positive and negative samples considering mu- sic emotion and phrase based on the public lyrics-melody paired dataset and divide songs to music pieces.Further,the lyrics and melody fragments are encoded by the lyrics-encoder and the melody-encoder,respectively.And take the en- coded lyrics feature and melody feature that are contextualized as the input of the lyrics-melody matching decoder to analyze the characteristic relationship between the lyrics and melody,and then determine the matching degree of the lyr- ics-melody segment.The experimental results show that the music lyrics-melody matching intelligent evaluation al- gorithm model based on sequence model can more accurately judge the matching degree of lyrics-melody matching than simple music emotion matching,which verifies the effectiveness of the proposed algorithm. Keywords:music lyrics-melody;emotion;rhythm;sequence model;lyrics encoder,melody encoder;matching decoder, lyrics-melody matching degree;music lyrics-melody matching 收稿日期：2020-01-06. 音乐词曲匹配度评估是针对音乐的词曲匹配 基金项目：国家自然科学基金面上项目(51775385,71371142)： 国家自然科学基金项目(71771176.61503287). 程度给出的一个客观评估。随着智能音乐研究发 通信作者：汪镭.E-mail:wanglei@tongji.edu.cn. 展，越来越多智能音乐作品产生，随之而来的便DOI: 10.11992/tis.202001006 基于序列模型的音乐词曲匹配度智能评估算法 陈壮豪1 ，张茂清1 ，郭为安2 ，康琦1 ，汪镭1 （1. 同济大学 电子与信息工程学院，上海 201804; 2. 同济大学 中德工程学院，上海 201804） 摘 要：情感匹配模型是一种常用于评价词曲匹配程度的方法；然而，单纯地依靠情感匹配模型无法对评价词 曲匹配度进行准确的评价。为改善此问题，提出了基于序列模型的词曲匹配度智能评估算法，其综合考虑词曲 情感和词曲间节奏关系以给出一个更加准确的词曲评估方法。基于公开词曲同步数据集，构建了音乐情感和 节奏正反例模型，并基于此模型将音乐切分成片段；进一步，将歌词和旋律片段分别通过歌词编码器和旋律编 码器进行编码，并将编码后具有上下语境的歌词特征和旋律特征输入词曲匹配解码器，解析词曲间特征关系， 判断词曲片段匹配程度。仿真结果表明：基于序列模型的词曲匹配度智能评估算法，相对于单纯的情感匹配模 型，能够更精确地评价词曲匹配程度，验证了本文提出算法的有效性。 关键词：音乐词曲；情感；节奏；序列模型；歌词编码器；旋律解码器；词曲匹配解码器；词曲匹配度； 中图分类号：TP393.04 文献标志码：A 文章编号：1673−4785(2020)01−0067−07 中文引用格式：陈壮豪, 张茂清, 郭为安, 等. 基于序列模型的音乐词曲匹配度智能评估算法 [J]. 智能系统学报, 2020, 15(1): 67–73. 英文引用格式：CHEN Zhuanghao, ZHANG Maoqing, GUO Weian, et al. Music lyrics-melody intelligent evaluation algorithm based on sequence model[J]. CAAI transactions on intelligent systems, 2020, 15(1): 67–73. Music lyrics-melody intelligent evaluation algorithm based on sequence model CHEN Zhuanghao1 ，ZHANG Maoqing1 ，GUO Weian2 ，KANG Qi1 ，WANG Lei1 (1. College of Electronics and Information Engineering, Tongji University, Shanghai 201804, China; 2. Sino-German College of Ap￾plied Sciences, Tongji University, Shanghai 201804, China) Abstract: Emotional matching model is a method often used to evaluate the degree of lyrics and melody matching. However, it cannot be accurately evaluated based on the emotion matching model. In order to improve it, this paper pro￾poses an intelligent evaluation algorithm of lyrics-melody matching based on a sequence model, which comprehens￾ively considers the emotion and the rhythm relationship between lyrics and melody to give an evaluation method for more accurate evaluation. Firstly, this paper researches and builds music positive and negative samples considering mu￾sic emotion and phrase based on the public lyrics-melody paired dataset and divide songs to music pieces. Further, the lyrics and melody fragments are encoded by the lyrics-encoder and the melody-encoder, respectively. And take the en￾coded lyrics feature and melody feature that are contextualized as the input of the lyrics-melody matching decoder to analyze the characteristic relationship between the lyrics and melody, and then determine the matching degree of the lyr￾ics-melody segment. The experimental results show that the music lyrics-melody matching intelligent evaluation al￾gorithm model based on sequence model can more accurately judge the matching degree of lyrics-melody matching than simple music emotion matching, which verifies the effectiveness of the proposed algorithm. Keywords: music lyrics-melody; emotion; rhythm; sequence model; lyrics encoder; melody encoder; matching decoder; lyrics-melody matching degree; music lyrics-melody matching 音乐词曲匹配度评估是针对音乐的词曲匹配 程度给出的一个客观评估。随着智能音乐研究发 展，越来越多智能音乐作品产生，随之而来的便 收稿日期：2020−01−06. 基金项目：国家自然科学基金面上项目 (51775385，71371142)； 国家自然科学基金项目 (71771176，61503287). 通信作者：汪镭. E-mail：wanglei@tongji.edu.cn. 第 15 卷第 1 期 智 能 系 统 学 报 Vol.15 No.1 2020 年 1 月 CAAI Transactions on Intelligent Systems Jan. 2020