Music and Audio Computing Lab

Symbolic Music Similarity


Main Contributors: Saebyul Park


Symbolic Melody Similariy

Melodic similarity is a key concept that helps with analysis and understanding in theoretical areas such as musicology, music cognition, and music psychology, as well as in applied areas such as music copyright, music classification and recommendation, and various fields. The process of determining melodic similarity is inherently intuitive and subjective. Psychological approaches to evaluating melodic similarity have relied on cognitive experimental evaluations, expertise, or music theory-based models, while computational methods derived from natural language processing have generally provided a single value. These fragmented approaches may only reveal task-specific information. Thus, our study aims to develop a comprehensive framework for quantitatively measuring the subjective and qualitative aspects of melodic similarity. To this end, we propose a novel method for evaluating the symbolic melodic similarity between two songs using multi-segment analysis of melodic sequences as well as visually inspecting cross-scape plots, which can compare two songs regardless of the location or length of similar phrases.

An Illustration of Steps for Drawing a Cross-scape Plot for Symbolic Melody Similarity

An example of a cross-scape plot by comparing similarities between two songs. The pixel value and the density of the color indicate the similarity score between the two sub-sequences.


The analysis procedure is summarized as follows: 1) Two songs are each segmented into all possible unit lengths, from the shortest to the longest. 2) A minimum local distance matrix between the two songs is constructed for all unit lengths. 3) The overall similarity measure is calculated by analyzing the cross-scape plots obtained by stacking up the local similarities. (Refer to the figures above.)

This hierarchical representation allows for capturing the location and length of similar segments between two songs in a visually intuitive manner. In our experimental research, we also showed the effectiveness of the cross-scape plot by evaluating it on examples from folk music collections with similarity-based categories and plagiarism cases. For further information, see the research paper linked below.


Related Publications

  • Mel2Word: A Text-based Melody Representation for Symbolic Music Analysis
    Saebyul Park, Eunjin Choi, Jeounghoon Kim and Juhan Nam
    Music and Science, 2024 [paper]
  • A Cross-Scape Plot Representation for Visualizing Symbolic Melodic Similarity
    Saebyul Park, Taegyun Kwon, Jongpil Lee, Jeounghoon Kim, and Juhan Nam
    Proceedings of the 20th International Society for Music Information Retrieval Conference (ISMIR), 2019 [paper] [code]