EMUSIC-L Digest Volume 40, Issue 16 This issue's topics: Grammars and Ragas (2 messages) Your EMUSIC-L Digest moderator is Joe McMahon . You may subscribe to EMUSIC-L by sending mail to listserv@american.edu with the line "SUB EMUSIC-L your name" as the text. The EMUSIC-L archive is a service of SunSite (sunsite.unc.edu) at the University of North Carolina. ------------------------------------------------------------------------ Date: Thu, 28 May 1992 14:38:33 CDT From: crispen Subject: Re: Grammars and Ragas Rajeev Upadhye writes: > The notion of "grammar" is widely accepted in Indian Classical music. >It is used to signify body of rules according to which a melodic fragment is >checked to see if it conforms to a particular "raag", without reference to >aesthetic quality of music in question. This notion has motivated us to look >into a more codified picture of raag music of India using the models of formal >languages in Chomskyan hierarchy. I presume that certain generative-grammar rules can also apply: that is, a composition is grammatical not only if the notes are within the given raga's (?) scale, but also if (for example) leading tones lead to their proper resolutions (that is, in the key of C, diatonic scale, a B-natural will be followed by a C-natural many times, and very seldom by an F-natural). You might even be able to have melodic "phrase structures". I'd be very interested in finding out what your rules are. Also what your morphemes are (see how long ago my linguistics classes were? I don't even think they're called morphemes any more). That is, is a B-natural bend to a C one element or two? >We have developed techniques to construct a computational model of raag music >of India based on the theory of finite state automata. Please excuse my ignorance, but doesn't a finite state machine imply a Markov process, and didn't Chomsky or Miller or somebody assert pretty definitively that a generative grammar can't be done via a Markov process? It's been many, many years, so if I'm all wet, please just say so and don't use up bandwidth trying to refute something I don't remember and probably didn't understand then. >If I perform a blind test, i.e., >if I record the music on a tape and play it to music lovers then they >fail to identify whether te composer is human or non-human. What about playing the samples to musicians? The criteria for "music lover" can be fairly elusive. Most musicians when they listen to a piece of music reconstruct the music in their heads, using at least as much cognition as emotion (you aren't going to catch *me* saying left-brain and right-brain). Active listeners may be able to tell the difference. Good luck to you on this project. Bob Crispen crispen@foxy.boeing.com ------------------------------ Date: Thu, 28 May 1992 14:43:47 EDT From: "Joseph D. McMahon" Subject: Grammars and Ragas music@parcom.ernet.in writes: > >From : Rajeev Upadhye > Email: music@parcom.ernet.in ... OR... > parcom!music@shakti.ernet.in > > Post : Centre for Development of Advanced Computing > Pune University Campus > Ganesh Khind > Pune, Maharashtra, 411007 > INDIA. > Telex: 0145-7615 CDAC IN > FAX : (0212) 337551 Dear Friends, I am taking this opportunity to introduce myself and my area of interest to the members of EMUSIC-L. I do not know whether by any chance any one of you got an opportunity to listen Indian Classical music. I am assuming that you are exposed to our music through artists such as Ravi Shankar and Yehudi Menuhin and several others, who did several experiments to popularise our music in the west. With this assumption it would be easier for me to explain what I am doing. The notion of "grammar" is widely accepted in Indian Classical music. It is used to signify body of rules according to which a melodic fragment is checked to see if it conforms to a particular "raag", without reference to aesthetic quality of music in question. This notion has motivated us to look into a more codified picture of raag music of India using the models of formal languages in Chomskyan hierarchy. Musical knowledge required for composing the raag music can be separated into two parts: raag-dependent & raag-independent. We have developed techniques to construct a computational model of raag music of India based on the theory of finite state automata. This model is able to generate valid note sequences in a given raag and also takes into consideration semantic aspects. It is possible to compare the grammars of two musical performances in the same raag objectively, in terms of commanality as well as differences. The music generated has received mixed response from music lovers ( most of the musicians are reacting violently to my ideas ). The music is generated through PC's speaker. If I perform a blind test, i.e., if I record the music on a tape and play it to music lovers then they fail to identify whether te composer is human or non-human. But these experiments are carried out using simple methodology. In future I expect the model to generate better and better musical ( melodic ) ideas which will assist musicians in music composition. >> So that my research is more product oriented.<< I am also planning to use Artificial Neural Nets to solve problems of pattern completion which are specific to Indian Classical music. I would be most happy to receive comments, criticism, ideas and from readers of this list. I would also like to answer any questions (to best of my capacity) by readers of this bulletin about the CLASSICAL MUSIC OF INDIA. with best wishes, Rajeev Upadhye ------------------------------ End of the EMUSIC-L Digest ******************************