Can computers learn the language of music? Prof Paritosh Pandya of TIFR is using his spare time to answer that question by creating a written script for Indian classical music, reports Deepa Narayanan

Prof Paritosh Pandya is a ‘left-brain’ kind of guy. Dean of the Faculty of Computer Science at the Tata Institute of Fundamental Research (TIFR) in Mumbai, mathematics and logic rule the better part of his day. But when he logs out of work in the evening, Prof Pandya flips a switch, as it were, and tunes into his intuitive and creative ‘right-brain’. In his quiet time, he revels in the strains of Indian classical music to recharge—but not in the way you think.

For many years, this scientist has been teaching computers the language of Indian classical music, a path-breaking effort and, arguably, the first of its kind. As if to explain, Prof Pandya points to his computer screen. It reveals row upon row of programming language, notations that could never be associated with the purity of a raga. Or so it would seem.

“Unlike Western music, which follows a script, Indian classical music is purely an oral tradition learnt at the feet of the master. But, ultimately, it is a language and one with a great deal of structure. With the use of concepts from programming languages, I am creating a script for Indian classical music called Swarupa,” says the scientist, whose area of work is building computer programs to mathematically analyse other programs for their reliability.

As his ‘hobby’, Prof Pandya first developed his Swarupa script and then used it to write a program for music synthesisers that recreates the sounds of instruments such as the tabla, santoor, tanpura, harmonium and flute. Still at a rudimentary stage, this software can provide instrumental accompaniment for vocal or instrumental performances and is a great tool to help beginners learn Indian classical music.

“Swarupa is a series of musical notations that I have developed over pre-existing but rudimentary notations by Bhatkande and Pusalkar, musicologists from the early 1900s who had written volumes on the language of Indian classical music,” reveals Prof Pandya. He explains that the very foundation of Indian classical music is vastly different from Western music. It is much more complex and nuanced, and therefore much more challenging to work with. While Western music goes by harmony, Indian classical music is big on melody and microtonal variations of each note, or the “shades of each note”.

To use Prof Pandya’s software program, one has to transfer it onto one’s computer from a CD or pen drive. “Each composition will have a different set of codes, depending on the sequence. The user creates one such file and feeds it to the Swarupa synthesiser, which will produce the accompaniment sounds of, say, the tabla or harmonium. Only a few home learners have tried it so far but they are very enthused,” says the scientist, whose journey with Swarupa began when he returned from a stint at the Oxford University Computing Laboratory in the UK.

“At Oxford, my curiosity about the use of technology in music grew. I started working on my own development in 1994-95. As a hobby, I worked on it consistently between 1996 and 2006; I was promoted to dean then [at TIFR] and my job required all my attention. Lately, I have returned to computing music.”

Clearly, Prof Pandya thrives on challenges and, with a touch of the poetic, remarks, “To create a language of Indian classical music is like attempting to create a language for painting. But with advances in computing and digital technology, we should be able to do so.”

Tracing his love for music to when he was a child, he recalls, “I think I can safely ‘blame’ it on two things in my life. One is my father who, though not a professional musician, was a keen classical music enthusiast and played the flute. So I woke up to his renditions every day, and music filled my home.” But what really got him hooked was an old Murphy radio set. “They bought it on my first birthday, when my family lived in a one-room apartment. Every day, I listened to the morning ragas until school time. When I returned home, I was back at it until 11 pm, sometimes sticking my ear to the radio set, listening to the maestros, so that I didn’t wake up the household.”

By the time he got to Class VII, he was reading magazines on electronics and trying to understand the workings of the radio. “For me, electronics and music came together. All through school, I tinkered with gadgets and had little projects. Soon, I knew exactly how music signals moved from one point in the radio to another. I began to understand radio waves and how small waveforms were amplified and delivered to the speaker,” he recollects.

He learnt to play the flute as a child and even trained in Hindustani vocals for a few years. When he was at M S University in Vadodara, which had one of the most prestigious music schools in the country, Pandya took classes from music students. And while doing his PhD at TIFR, he learnt music under the scholarly guidance of Namdev Panchal, a lab technician who had trained under some of the masters of Indian classical music. “That’s when I really understood the intricacies of Indian classical music and ragas, and got a sense of what to look for in music.”

It’s been a long journey since then—but it’s still only the beginning. “My work so far has to do with the ‘synthesis’ of music, or producing sound from text. Then there’s another theme, which is ‘analysis’, or producing text from sound,” he reveals. Breaking it down, Prof Pandya says that it would be possible to create a software program to convert a musical rendition into a written script, all on its own. Using an analogy, he says, “Take a speech analysis program—the kind you use in Dictaphones—that can transcribe what you say by analysing your accent. It’s transcription software that understands the rules of grammar. Similarly, we could use computers to script a vocal or musical performance in real time. The uses for this extend beyond just learning music.”

Gazing into the future, Prof Pandya says, “This is where ‘artificial intelligence’ will come into play in a big way. It will be used to analyse Indian classical music, to identify patterns in various compositions and ragas, to see how the same set of notes is used in different ways. This is the same data analysis technology used to detect credit card fraud, gene sequencing and, in recent times, big data. Essentially, there is a whole lot of data and no formalised rule for how most things take place. Yet there is a pattern that these techniques will be able to recognise and extrapolate, and predict.”

It is rare to find a computer scientist standing on the threshold of a breakthrough in the classical music field. The excitement in this development was evident recently at the Bhau Daji Lad Museum, where Dr Pandya delivered a lecture, as part of the Junoon Theatre’s Mumbai Local workshops.

“I believe the division between arts and science is artificial,” he is quick to point out. “Life is a continuum and you can place yourself wherever you want. In my work, I deal more with the science side of things but I keep coming back to the arts to recharge my batteries.”

Photograph by Sumiron Ghosh / Junoon Theatre-Mumbai Local
Featured in Harmony — Celebrate Age Magazine
November 2016