After a year-long research and development of various tools to aid his goal, young lyricist and computer science researcher Madhan Karky says he is close to coming up with a computer-automated solution that will help Tamil film lyricists generate songs based on stories and tunes.
“Unlike a poem (kavithai) that comes from the soul of a poet, Tamil lyrics are written to suit situations in stories and are sometimes set to the tunes which the music director provides. There are also unique demands from producers as they want some catchy meaningless words or English words,” Mr. Karky, son of veteran lyricist Vairamuthu, says. “This is not very different from having a framework as per the requirements of a client and coding software to provide specific results.”
Some software tools that will help Mr. Karky realise his goal of ‘lyric engineering’ have already been published as free tools on the portal www.karky.in/labs by his not-for-profit research organisation, Karky Research Foundation.
One such tool, Paadal, is a lyrics search engine that throws up results from a database of over 2,500 Tamil songs from movies dating back to 1940s. (A keyword search for ‘Kaadal’ (love), for instance, lists nearly 100 songs.)
Another tool, Emoni, is a rhyme-finder that allows users to type in a Tamil word and find rhyming words. The lyricist used this tool while writings the lyrics for his popular song ‘Aska Laska’ in director Shankar's ‘Nanban’.
Mr. Karky is being helped in his project by two full-time staff at his Foundation, and a group of dedicated volunteers, including engineering college students. He had resigned as an assistant professor from the department of computer science at Anna University earlier this year and supports the research entirely through earnings from his film career as a lyricist and dialogues writer.
There are more complex building blocks to Mr. Karky’s project, some of which are in various stages of conception and realisation.
A complex challenge is a tool that will help construct new Tamil similies based on word associations. In its beta testing stage, the algorithm turned one up: ‘A relationship that gives joy like a festival’ (Thiruvizha pola santhosham tharum uravu).
“That was our first machine-generated simile. When I tweeted it a few months back, and teased people to guess whose lyrics it was, people started saying it sounded like Vaali or my father’s song,” says Mr. Karky.
Once completed, the project will be able to simplify the process of writing lyrics. Users can generate a song by feeding in basic details, such as the genre of the song (character, philosophy, romance and so on), its mood (happy, sad, angry) and style (classical, folk, multi-lingual or mixed).
“The idea is not to replace the lyricist with a machine,” he says, “But to help him with technology.”