《九宫大成南北词宫谱》工尺谱数据集构建及其若干应用
摘 要
:
本研究对《九宫大成南北词宫谱》及其译本进行了数字化,构建了包含工尺谱、唱词与相应简谱翻译对齐的、多源异构的中国传统音乐数据集。运用700余首人工录入所得工尺谱与简谱的小样本数据集设计并训练出基于深度学习的解谱模型,由此翻译了余下约5800首曲目,最终整理得到含6563首曲目的全本数据集。此外,还在数据集上进行了若干应用探究:进行音乐形态学分析,探讨了谱字声调与唱腔动态间的关系;建立了数字化乐谱的在线展示平台。本研究意在填补音乐计算领域中关于中国传统音乐的数据空白,并为机器理解传统音乐提供一些新的思路。
关键词
:
《九宫大成》;工尺谱;音乐数据集;乐谱数字化;古谱翻译
文章编号
:
1002-9923(2024)04-0129-09
DOI:
10.13812/j.cnki.cn11-1379/j.2024.04.014
作者简介:
李荣锋(1984— ),
男,汉族,博士,北京邮电大学讲师。
卜 凡(2001— ),
男,汉族,北京邮电大学2020级本科生。
苏倩澜(2000— ),
女,汉族,北京邮电大学2018级本科生。
基金项目:
本文为2019年度教育部人文社科青年基金项目“基于机器学习的中国工尺谱自动翻译研究”(项目编号:19YJCZH084)的阶段成果。
Construction and Applications of the Gongche Notation Dataset in
Jiugong Dacheng Nanbei Ci Gongpu
○Li Rong Feng, Pu Fan, Su Qian Lan
Abstract:
This research involves the digital processing of traditional Chinese music sheets from the
Jiugong Dacheng Nanbei Ci Gongpu
and their translations, resulting in the construction of a multi-source, heterogeneous dataset of traditional Chinese music. This dataset contains both Gongche notation and lyrics, along with their corresponding translations. Initially, a small sample dataset was manually created for over 700 songs, which was then used to train a deep learning-based translation model. This model was subsequently applied to translate approximately 5800 additional songs, culminating in a comprehensive dataset of 6563 songs. Several exploratory applications were conducted using this dataset, including morphological analysis to examine the relationship between the tones of the lyrics and the dynamics of the melodies. Additionally, an online display platform for the digitized songs was established. This study aims to bridge the gap between the fields of music computation and traditional Chinese music, offering new insights for the application of machine learning in the understanding of traditional music.
Keywords:
Jiugong Dacheng
; Gongche Notation; Music Dataset; Music Sheet Digitization; Traditional Musical Score Translation
本文刊登于
《中国音乐》2024年
第4期
专栏:晚清近代工尺乐谱研究
第129-137页(总第188期)
【照片素材,版权属于原作者,若涉及版权问题,请原作者联系我们删除处理,我们只做分享,不用于商业。】