Posters

Eunwoo Heo (POSTECH)

Title: Applications of topological data analysis to the featured graph obtained from time series data

Abstract: In this research we provide a novel method that applies topological data analysis (TDA) to the featured graph obtained from the time series data. In addition, we show that the proposed method is stable. For the featured graph, we use the time series data of music. In our earlier study, we considered Korean music as the target time series data. From the target time series data, we constitute the music network by defining the nodes and edges with an appropriate distance between the nodes and the weights of nodes and edges. With the defined distance and weights, the Vietoris-Rips filtration is applied for TDA. However, the previous study did not fully consider the unique features of Korean music e.g., “Sikimse” and “Jangsik-eum” (ornamented notes). To take them into account in TDA, we classify each and assign the type values to each. Then we propose a novel method that utilizes the variable distances between the nodes. Contrary to the previous method for which the distance between edges is fixed regardless of the existence of Sikimse or Jangsik-eum, the new method allows the distance to vary depending on those feature values. By varying the distance, we can analyze how those unique features by Sikimse and Jangsik-eum can change the geometric structure of the given music. We apply the new method to Korean music written in Jeongganbo. We also proved the stability theorem of the proposed method applied to the featured graphs. Since the procedure of the proposed method is general, it can be applied to various similar graphs from time series data.

Hyo Jung Huh (Chosun University)

Title: Ideological Construction of ‘Sublime’ in German-speaking Music Discourse around 1800

Abstract: This paper traces the semantic change of 'sublime' that occurred in German-speaking music discourse around 1800, and argues that it was an ideological construction that occurred in the process of matching Kant's concept of the sublime with their musical idealism by contemporary music critics. In the German-speaking music discourse in the mid-to-late 18th century, ‘sublime’ was a term that referred to specific musical characteristics such as simplicity, pathos, and holiness, and gave positive connotations to related musical works or styles. However, 'sublime' did not mean that it was superior to musical works with other characteristics. On the other hand, around 1800, the term 'sublime' in music discourse began to mean 'absolute excellence' and represent canonical status. This paper examines that the scope and the status of the musical sublime had been changed by connecting the subjective, supersensible, and moral aspects of Kant's notion of the sublime with specific musical elements, and it led to the understanding of sublime as 'absolute excellence' in music discourse. In Kant's theory, the sublime belongings to the subject, and it is hard to apply to the object of art. In addition, Kant’s notion of sublime conflicts with the conventional nature of music at that time in that it presupposes the anti-purpose of the object. However, music critics of the time applied Kant's sublime to musical art and connected the subjective aspect of the sublime to the musical properties (the objective). In this process, a logical leap occurs between the general definition of the sublime and the expression of the sublime in music. In order to overcome this self-contradictory logic, contemporary music critics modified 'sublime musical characteristics' and expanded the category of the sublime. In addition, they modified the supersensible and moral aspects of Kant's sublime into the musical idealism of the time and the historical concept of the sublime, and constructed the new meaning of musical sublime with metaphysical and anti-secular connotation. It led to changes in the scope and the status of sublime in musical discourse, and ‘sublime’ became a term representing true art and began to mean ‘absolute excellence’ in the music discourse. 

Hyunjae Kim (KAIST)

Title: Human Cognitive evaluation for interpolational AI-composition based on VAE models.

Abstract: In recent years, AI-based music generation models have increasingly been utilized to assist human creativity. Accordingly, the issue of originality in AI-generated music has emerged as a significant concern, closely related to how humans perceive the musical similarity of melodies. Despite the importance of the issue, there has been limited research on the topic, particularly in the context of an in-depth evaluation with respect to human perception of music generated by AI models. In this study, we first investigated the relations between AI-composition models and human composers by projecting human-composed interpolation melodies, into the latent spaces of several VAE-based composition models. The results suggest how AI models are differently aligned with human interpolation. Second, we examined how humans perceive the musical similarity between interpolated melodies generated by an AI model. We conducted a human behavioral experiment, in which 32 participants were asked to rate the perceived similarity between interpolated melodies. The experiment shows that perceived musical similarity is linearly correlated with the level of interpolation by an AI model. With the results, we provide the initial insights for evaluating AI-composition models from the perspective of human cognition, and explore potential directions of AI-composition for future research.

Jiwon Kim, Jisu Lee, Jihyun Jun, Yewon Yun, Yeojin Kim, Jung Ah Lee, Juyong Uhm, Subin Park, & Jeong Mi Park (Seoul National University)

Title: Exploring the Relationships between Dialects and Folk Songs

Abstract: This study investigates the influence of language prosody on musical composition. While musicologists and linguists have recognized the impact of prosody on musical structures across different countries over time, there is still limited exploration of such influences within regional dialects of the same country. Therefore, our research aims to fill this gap by examining how pitch and rhythm in folk songs correspond to the distinct dialects of two regions in South Korea: Gyeonggi province, which utilizes Standard Korean, and Gyeongsang province, known for its unique dialect. We compiled folk songs ("Minyo") from both regions and scrutinized the variations in pitch and rhythm across each dialect and folk song. The pitch intervals, classified as stepwise motion or leap, were computed into a 'leap-interval ratio' and compared using independent t-tests. Rhythmic variability on the other hand, was assessed via the normalized Pairwise Variability Index (nPVI) and analyzed through Mann-Whitney U-tests. Our results indicated a significant preference for leap intervals in Gyeongsang region's folk songs compared to those from Gyeonggi region (t(19) = -3.325, p = 0.004). This trend for leap intervals was also consistently seen across the dialects of the two regions, with the Gyeongsang dialect exhibiting a higher leap-interval ratio than Standard Korean (t(19) = -2.249, p = 0.044). Conversely, we found no statistically significant differences in rhythm between the songs or the dialects of the two regions. These findings lend empirical support to the concept that a region's spoken language prosody shapes the structural composition of its folk music, thereby suggesting that a region's dialect can imprint distinctive melodic characteristics onto its music, even within different regions of the same country. 

Eunji Oh (KAIST)

Title: Exploring the influence of online music learning on linguistic, cognitive, and musical skills in Korean preschoolers: A pilot study

Abstract: Music training has a range of positive effects on early childhood development. Recent research has demonstrated that music interventions can improve language and cognitive abilities, particularly benefiting preschool children. However, it remains uncertain whether online music learning is as effective as in-person learning in terms of children's abilities. This study aimed to investigate the impact of online music training on typically developing children. A music program consisting of 12 sessions was conducted using the Zoom platform, with the participation of twenty-one 5-6-year-old Korean children. Prior to and after the training sessions, assessments were conducted to measure reading, cognitive, and musical abilities. The results showed that online music education enhances all three skills in preschool children. This study highlights the promising potential of remote music learning in facilitating the abilities necessary for children's development.

Eun Ji Park (Seoul National University)

Title: A Study on Automatic Music Composition Methodology in the 17th Century: Focusing on Kircher's ‘Arca Musarithmica’

Abstract: Athanasius Kircher (1601–1680), a 17th-century German-born Jesuit priest, theorist, and scientist, published Musurgia Universalis in 1650. This book is an overview of drawing on mathematical theories from ancient and contemporary music and dealing with them along with health and rhetoric. In this book, Kircher proposed Arca Musarithmica, a device that uses combinatorics to help the process of composition. ‘Arca Musarithmica’ was created based on the musical thinking of Kircher, who had extensive knowledge of linguistics in the past, and can be seen as one of the first automatic music composition devices. Through this device, Kircher devised a method with which even those who do not know music can compose easily. ‘Arca Musarithmica’ could create millions of four-part music through a simple combinatorial technique. Moreover, the device also has notable historical value in that it lays the foundation for today's computing technology, AI music methodology. This study organizes the musical principles of ‘Arca Musarithmica’, which appeared in Kircher's book Musurgia Universalis, and illuminates and analyzes the value by following its historical traces.