A lyricist in Tamil film music, by popular definition, is a writer with intensity, who longingly awaits his muse from the comfort of his home, for the magical inspiration that gives birth to a five-minute-long hit.
Hard to believe then, that Madhan Karky—for whom writing lyrics is akin to an optimisation problem in Maths— could be a successful lyricist. As if to mock the stereotype, he even practices his profession from a bustling coffee shop in Besant Nagar, Chennai.
The secret to his success— he claims to have perfected a software for generating words for any given tune in Tamil.
He explains the origins of his inspiration—his lack of proficiency in Tamil. “My knowledge of Tamil is confined to film songs, and about 40 books,” he says with characteristic calm and an unassuming tone. It didn’t help that Vairamuthu, his father, is one of Tollywood’s most respected lyricists, while mother Dr Ponmani is a retired professor of Tamil. Karky himself is, in fact, listed as an assistant professor, at the Tamil computing lab of Anna University.
“Both personal exigency and a need for variety in the lyrics I wrote, forced me to develop this system,” says the Australia- educated software engineer. “Lyric-writing doesn’t involve strict formats, yet, I noticed our lyrics used some words again and again. Given my poor Tamil, I needed help to write songs,” he elaborates. The 25-songold lyricist conjured up his system while writing his first hit, Irumbile oru irudhayam for the sci-fi Enthiran.
The system is simple.
Once Madhan gets a tune from a composer, he uses a software to split it across a template.
Once broken down, he writes to the small parts, keeping the required emotion in mind. When he runs out of words or needs to fill gaps, he resorts to his invention, feeding it information on the sentence—its length, rhythm, etc— which then generates words to fit the bill.
“I can bring in more sensitivity in my search, by specifying words that rhyme, alliterate or are synonymous to another word,” he explains. If unhappy with the result, the system provides alternatives that also rhymes with the first, or as need may be, shares its first letter and still fits the song’s rhythm.
Since ascertaining the freshness of a word is important, he added a freshness score, which ranks words based on how often it has been used on the Internet.
The vocabulary is sourced from agaraadhi, an online Tamil dictionary with over three lakh words, developed by Madhan’s lab. “The software includes parameters to measure the pleasantness of a word, i.e. how good it sounds when uttered,” he adds.
Madhan’s software also does pattern analysis for deeper understanding of hits from the past, and to break down current trends in lyrics. “If a line ends with bhoomi, odds are, the next line would end with saami,” he explains. The 30-yearold engineer is currently working on Wordnet, a project to suggest related words in Tamil.
Ask him if such systems kill creativity, he retorts: “There are electronic influences everywhere. Voices of singers are smoothened with autotune software in post-production. My system uses science to support art.”