Fig 1.
Constructing the network of artists and cultural products.
(A) The comprehensive classical music recordings data from ArkivMusic is a bipartite network with edges running between CDs and the musicians. The musician layer (bottom) is a heterogeneous mix of musician classes—composers, conductors, ensembles, and individual performers. (B) A backbone of the network of musicians (CDs omitted via one-mode projection). An edge between musicians means that their compositions or performances were featured on a common CD.
Fig 2.
The multiscale views of the network landscape of the classical music network.
On the macroscopic level (top), we take a bird’s-eye view of the global characteristics of the network. On the mesoscopic scale (middle), we investigate the community structure of the network that reveals the homophily based on musician characteristics such as period and nationality. On the microscopic scale (bottom), we find the local network landscape around a specific musician by quantifying the relevance of others to the musician. This type of multiscale view allows us to correctly characterize the relationships between musicians and their roles in the cultural collaboration network, where a simple global prominence (top) can easily eclipse the rich local structures that represent diverse styles (middle) and individuality of artists (bottom).
Table 1.
Basic Network Properties.
Fig 3.
Measuring the macroscopic network structure.
On the macroscopic scale the network is characterized by wide variations in the visibility of the musicians, potentially masking diversity and the existence of smaller structures. (A) The cumulative degree distribution P(K > k) of nodes in the bipartite network. It appears to follow a power law P(K > k) ∝ k−τ+1 (with τ = 2.31 ± 0.03), suggesting an extreme level of difference in the visibility or prominence between musicians. (B) Composers from the Romantic period are overrepresented in the lists of highest-degree composers. For instance, nearly 50% of 100 highest-degree composers are from the Romantic period (far left), while it accounts for only 10% of all composers (far right). (C) Significant variations in the degree of musicians are observed within the musical period as well.
Table 2.
Ten Highest-Degree Musicians (Composer, Performer, Conductor, Ensemble).
Fig 4.
Communities showing the mesoscopic network structure.
On the mesoscopic scale the network is characterized by tightly-knit communities. We show four major communities. We show which musician attributes (composer periods, performer positions, and musician nationalities) are overrepresented in each community. Community A, for instance, represents the Austrian-German Romantic music; B represents the USA-based Modern music; C represents the transitional period between Romantic and Post-Romantic; finally, D represents the classical guitar.
Fig 5.
Microscopic network structure centered on an individual musician.
(A) The musicians most relevant to violinist Kyung-Wha Chung as an ego determined by Egocentric PageRank (EP, left) and Degree-Neutralized Egocentric PageRank (DNEP, right). Of the ten highest-EP musicians, seven (Tchaikovsky, London Symphony, J. S. Bach, Beethoven, Royal Philharmonic, W. A. Mozart, and Brahms) are also among the ten highest-degree nodes in the overall network. The ten highest-DNEP musicians feature those more specific to the ego, with the Chung Trio (composed of Chung’s two siblings) occupying the top spot, with Krystian Zimerman and Simon Rattle known for their collaborations with Chung in high spots. W. A. Mozart, in contrast, falls rapidly in the ranks. (B) A figure showing the egocentric network landscape determined by DNEP (β = 1) around Kyung-Wha Chung. Highly relevant musicians tend to be lower in degree but more specifically related to her (e.g., composer Max Bruch, conductor Charles Dutoit, Montreal Symphony Orchestra, etc.). High-degree nodes such as Tchaikovsky are pushed outwards, demonstrating the ability of DNEP to differentiate between ego-specific and universally associated musicians.
Fig 6.
Versatility of composers based on relevance to instrument groups.
The number of composers highly relevant (ranked 100th or higher) to any of the five largest instrument groups (violin, cello, piano, tenor, and soprano) is in the circles. Composers relevant to multiple instrument groups in the absence of degree effect tend to be the universally recognized composers, revealing the connection between macroscopic and microscopic network patterns.