title

 

Inter-Composer Similarities Statistics

In the attached pdfs two sets of summary statistics are given for the 'Most Similar Composers' lists that recently (9/2019) have been added to the "Composers" section on a composer-by-composer basis. In that section each of the five hundred subject composers have been assigned a list of those fifteen composers from that group who are most similar to them, based on a statistical analysis. Following are a description of the derivation of the similarities, and some comments on how the information might be used.

Derivation

Practically speaking, it is useful to recognize similar composers simply because one might wish to know of other individuals whose music resembles a person already known, and liked. But the notion of 'similarity' can be extended to further considerations, including ones reaching beyond the music itself. Here, I have incorporated all the data collected, including composer-specific information on influenced and influencing composers, geographical associations, favored instruments and genres, historical periods, etc., to produce 'fields' of variables that could be examined for overall correlations between particular pairings of composers. The centralized cosine index was the statistical analysis device employed; this provides a score (ranging from 1.0 to -1.0) for the degree of matching between these 'fields' for any pairing of composers (see Smith et al. 2015 in the bibliography below). To these scores can be attached statistical significance levels, but for the layperson the scores themselves are easier to appreciate: generally speaking, scores above .60 represent composer similarities that are likely to be fairly obvious, scores of about .45 to .60 signify a considerable similarity, .30 to .45 some similariy (for example, of time period and emphasis on guitar), and below .30 less obvious connections (though many of these may be statistically significant in the greater sense).

In the "Composers" section the fifteen highest scores associated with an individual are given in the final field for that entry. I have also compiled a list of these, arranged according to composer ranks (where J. S. Bach is #1, etc.):

retrieve PDF file here

Application

It must be re-emphasized that the similarities scores arranged here represent appraisals of correlations between pairings of composers' recorded attributes. This means not only instances in which two composers share the same attribute, but also the extent of 'unmatches'--that is, where one of the two has many more attributes recorded than the other. JS Bach, for example, has many more recorded attributes here than does Johann Ludwig Krebs, a relatively minor figure greatly influenced by Bach. The result is that Bach shows up as eighth most similar composer in the Krebs entry, but Krebs does not appear in the Bach list.

Neverthless, the lists do pass, at least largely, an eye test. Plus, they lend themselves to a related notion, that of 'representativeness.' Usually when one speaks of a 'representative' composer a famous person is pointed to: for example Brahms for the Romantic Period, or Debussy for Impressionism. These are, however, major geniuses and are perhaps not really representative of their period or school. Here, one can get a different perspective by seeking out those composers who appear on the 'top fifteen' lists the most, or least, number of times. Those appearing the most times can be thought of as being highly representative, or most typical; those appearing the fewest times can be thought of as being least representative: that is, they are some combination of being most atypical, distinct, or unusual. I have compiled ranked lists of each condition:

retrieve PDF file here

Bibliography of CMN-related Publications

--Patrick Georges & Ngoc Nguyen, 8 April 2022. "Visualizing music similarity: clustering and mapping 500 classical music composers." Digital Humanities Research 2(1): 68-85. [in Chinese; transl. Zhang Jiaming]

--Patrick Georges & Aylin Seckin, 24 February 2022. "Music information visualization and classical composers discovery: an application of network graphs, multidimensional scaling, and support vector machines." Scientometrics. https://doi.org/10.1007/s11192-022-04331-8

--Patrick Georges, September 2019. "Visualizing music similarity: clustering and mapping 500 classical music composers." Scientometrics 120(3): 975-1003.

--Patrick Georges & Ngoc Nguyen, July 2017. "Western classical music development: a statistical analysis of composers similarity, differentiation and evolution." Scientometrics 112(1): 21-53.

--Charles H. Smith & Patrick Georges, 2015. "Similarity indices for 500 classical music composers: inferences from personal musical influences and 'ecological' measures." Empirical Studies of the Arts 33(1): 61-94.

--Charles H. Smith & Patrick Georges, 2014. "Composer similarities through 'The Classical Music Navigator': similarity inference from composer influences." Empirical Studies of the Arts 32(2): 205-229.

--Charles H. Smith, Patrick Georges & Ngoc Nguyen, December 2015. "Statistical tests for 'related records' search results." Scientometrics 105(3): 1665-1677.


Return to Main Stats Menu
Return to Main