Visualizing Symbolic Music via Textualization: An Empirical Study on Chinese Traditional Folk Music

The continuous integration of computer technology and art has promoted the development of digital music. In recent years, the massive music data has promoted the problems of music inquiry, music classification, music content understanding and so on, making a new subject develop continuously, that is music information retrieval. The current mainstream feature extraction of music is based on acoustics, such as pitch, timbre, loudness, zero-crossing rate, etc. Direct symbolic music feature extraction and music analysis are relatively rare. This paper aims to present an empirical study on text clustering ideas in natural language processing into the field of symbolic music style analysis. Firstly, Symbolic music is textualized and proposed by us as an inspiration. To be precise, textualization of symbolic music is converted into weighted structured data through tf-idf algorithm. In the following step, three different types of mainstream clustering algorithms, K-Means, OPTICS, and Birch are used to perform cluster analysis and comparison on the traditional Chinese folk music dataset we crawled, T-SNE algorithm is used to visualize the dimensionality reduction of high-dimensional data. Finally, a series of objective evaluation indicators of clustering are used to evaluate the three clustering algorithms. Through comprehensive evaluation of indicators, it is proved that the clustering algorithm has achieved an excellent clustering effect on the midi note dataset we extracted. As a result of the clustering, the professional music theory knowledge and the historical development characteristics of traditional Chinese folk music are comprehensively integrated, which reversely verifies that the 1300 midi music data sets have distinct modal characteristics of traditional Chinese folk music.