The Role of Statistics in Computational Musicology

• Published in: Computational Musicology in Hindustani Music pp. 15 - 24
• Main Authors: Tewari, Swarima, Mazzola, Guerino, Chakraborty, Soubhik, Patra, Moujhuri
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2014; Springer International Publishing
• Series: Computational Music Science
• Subjects: Applied mathematics; HUMANITIES; MUSIC; Musical Data; Musical Performance; Musical Piece; Musical Structure; Revealing Feature
• ISBN: 9783319114712; 3319114719
• ISSN: 1868-0305; 1868-0313
• DOI: 10.1007/978-3-319-11472-9_2

Music analysis, broadly speaking, can be divided into commonality analysis (“what is common?”) and diversity analysis (“what is special?”). It is worth pointing out here that a statistician is essentially a commonality expert in the sense that the philosophy of statistics is to summarize and average and make inferences which are true on the whole and which describe a process rather than an individual entity. Fortunately, there are issues in music where this traditional mindset of the statistician finds a support. For example, a collection of recordings of the same artist if analyzed statistically will definitely reflect certain common features having to do with the style of the artist. But the statistician must realize that every single music piece will have something special to offer. Fortunately, again, there are issues even in statistics where the statistician does take an individual observation seriously—as in the case of an outlier or influential observation, for example. There is a whole literature in statistics to deal with outliers. When an outlier comes, the traditional philosophy of summarizing and averaging is brushed aside. The statistician goes after this individual influential observation exploring how it came and what it signifies. The case of outliers is an exception in statistics. It is the very grammar in music as music is a work of art!If the statistician can use his experience and mindset of handling outliers (regarding every musical piece as a musical outlier) along with his commonality expertise, he can be a very effective music analyst. Similar point of view has been expressed by Nettheim who has also provided a good bibliography of statistical applications in musicology (Nettheim 1997). For a sound statistical treatment of musical data, see Beran and Mazzola (1999). Additionally, we acknowledge the contributions from Meyer (1989), Snyder (1990), Winkler (1971), Wilson (1982), Todd and Loy (1991), and Morehen (1981). It is an irony that computational musicology in Indian classical music is still lagging behind the progress in Western classical counterpart, although we do appreciate the efforts of Castellano et al. (1984), Chordia and Rae (2007), and Sinha (2008) among others. We hope this book will provide some food for thought in that direction. Statistics is a useful tool both for analyzing a musical structure and quantitative assessment of a musical performance. The former helps in revealing features of a musical piece in general, while the latter brings the style of the artist into consideration as well in performing the musical piece.