In this study we introduce a symbolic dataset composed of non-metric Iranian classical music, and algorithms for structural parsing of this music, and generation of variations. The corpus comprises MIDI files and data sheets of Dastgah Shour from Radif Mirza Abdollah, the foundational repertoire of Iranian classical music. Furthermore, we apply our previously-introduced algorithm for parsing melodic structure (Kanani et al., 2023b)to the dataset. Unlike much Western music, this type of non-metric music does not follow bar-centric organisation. The non-metric organisation can be captured well by our parsing algorithm. We parse each tune (Gusheh) into a grammar to identify motifs and phrases. These grammar representations can be useful for educational and ethnomusicological purposes. We also further develop a previously-introduced method of creating melodic variations (Kanani et al., 2023b). After parsing an existing tune to produce a grammar, by applying mutations to this grammar, we generate a new grammar. Expanding this new version yields a variation of the original tune. Variations are assessed by a domain-expert listener. Additionally, we conduct a statistical analysis of mutation with different representation setups for our parsing and generation algorithms. The overarching conclusion is that the system successfully produces acceptable variations post-mutation. While our case study focuses on Iranian classical music, the methodology can be adapted for Arabic or Turkish classical music.

翻译：本研究引入了一个由伊朗非节拍古典音乐组成的符号数据集，以及用于该音乐结构解析与变奏生成的算法。语料库包含伊朗古典音乐基础曲目《拉迪夫·米尔扎·阿卜杜拉》中“舒尔达斯特加赫”的MIDI文件与数据表。此外，我们将先前提出的旋律结构解析算法（Kanani等人，2023b）应用于该数据集。与多数西方音乐不同，此类非节拍音乐不遵循以小节为中心的组织形式，而我们的解析算法能有效捕捉其非节拍结构特征。我们将每首曲调（古谢）解析为语法结构以识别动机与乐句，此类语法表征可用于音乐教育与民族音乐学研究。我们进一步拓展了先前提出的旋律变奏生成方法（Kanani等人，2023b）：通过解析现有曲调生成语法结构，对该语法实施变异操作后产生新语法，扩展新语法即可生成原曲调的变奏版本。变奏成果由领域专家进行听觉评估。此外，我们针对解析与生成算法的不同表征设置进行了变异操作的统计分析。总体结论表明，该系统能成功生成经变异处理后可被接受的变奏。虽然本案例研究聚焦伊朗古典音乐，但该方法论可适配于阿拉伯或土耳其古典音乐研究。