A key concept in flamenco singing is the constant re-interpretation of orally transmitted melodic templates, which undergo variation and ornamentation during spontaneous performances. We propose a novel framework which allows us to model a set of interpretations of the same melody and analyse differences among them. We apply a multiple progressive sequence alignment technique to obtain a graphical model holding the frequency of occurrence of characteristic notes and note transitions, from which the underlying melodic template can be approximated. In a case study, we showcase how visualisations of the derived model yield insights into local and global melodic stability and inter-style contour similarities. We furthermore show how the extracted templates can used to identify instances of a particular melody in a large corpus of automatic transcriptions.

弗拉门戈舞的关键概念演唱是对口头传递的旋律模板的不断重新诠释，该模板在自发表演中会经历变化和修饰。我们提出了一个新颖的框架，该框架使我们能够对一组相同旋律的解释进行建模，并分析它们之间的差异。我们应用多重渐进序列比对技术来获得一个图形模型，该图形模型保存了特征音符和音符转换的发生频率，从中可以近似基础旋律模板。在案例研究中，我们展示了派生模型的可视化如何产生对局部和全局旋律稳定性以及样式间轮廓相似性的见解。我们进一步展示了提取的模板如何用于识别大型自动转录语料库中特定旋律的实例。