Jalin Gladis and Ganga Kameswaran, both students of computer science at College of Engineering, Guindy, have been avid fans of Tamil film music for as long as they can remember. Over the last three years though, their involvement with this medium has touched another level. Now they listen to almost 300 songs on a particular ‘emotion,' analyse each one to break it down in every way possible.

Nostalgia, betrayal, love, friendship, melancholy, sadness – the list is endless, but the quest is still on to find the driving emotion behind every song. And all this is part of a project – to automate the process of writing Tamil lyrics, one of the most painstaking bits of work behind a song.

“They asked us to do something in the area of music. We decided to create poetry,” says Ganga. Final year projects for computer science students involve many programming skills, but for a few, such as them, it is also about delving deep into their passion. Around 20 students from computer science and Information Technology, guided by a staff member, are working in this area of lyric generation.

With the demand for musicians rising to an all-time high, the question of what the contours of a ‘new' creativity will be is a question these young engineering students attempt to tackle with technology.

It is all about looking at lyrics from a statistical point of view and trying to find the patterns of creativity, say these students. “We have created a corpus of words for every emotion,” says Jalin. So ‘Idhayam Padhariyadhu,' ‘Jeevan Vaadiyadhu' and ‘Siluvayil araindhaai' would fall under betrayal, while a whole lot of other words and phrases fall under, ‘Sogam,' and ‘Anbu.'

There are popular similes and rhymes added to make the set of patterns wider, apart from, details of syntax and grammar. The software then, through random searches and probabilistic methods, creates lines depending on the ‘situation' described. It uses semantics (rules for interpreting the syntax) and so the challenge here is to study the pattern of every song and get as many words into the corpus, say the students.

Much of what these students are into – lyric engineering - is inspired by Madhan Karky, son of Tamil lyricist Vairamuthu who uses mathematical algorithms to find patterns of words. Mr. Karky, an assistant Professor at Anna University, also a lyricist who has worked in films such as ‘Endhiran,' says “When a person writes poetry, it is all his creation. But lyric writing is all about engineering, fitting words in a preset tune in a preconceived situation.”.

Such tools will only make the life of lyricists easy, and in fact, will also encourage more to write, he feels.

“Many people have ideas but less control over language. These tools are like formulae and spreadsheets that made the accounting process simple,” he says. Some of the tools he has developed include, one that finds out the words least used in Tamil songs, and those most used.

“We need to stop using the computer as a rule based machine. Wouldn't it be nice if the system gives simile never used before,” he asks.

But, will the listeners accept such lyrics? “We think they would. In fact, they might not even know it is machine-generated. After all, poetry means different things to different people,” says Ganga.


Vasudha VenugopalJune 28, 2012