Every time I crack open a new book, I inevitably end up listening to music while I read, subsequently forming an association between the text and the tunes. Sometimes the pairing doesn’t match (Swans and Murakami—nope), but ultimately the two pieces of art end up forming a bond where experiencing one by itself triggers thoughts about the other. Still, I often wonder what the souls of my favorite books would look/sound/smell like—what they would look like if turned into a painting or translated into cohesive album. A new program called TransProse (clever name, right?) does just that, as it uses a nuanced system to transform key characteristics from certain books into music based on those qualities.
Created by programmer and artist Hannah Davis and Saif Mohammad, a research officer at the National Research Council Canada, TransProse strives to compose scores out of the heart and essence of various novels, including Peter Pan, A Clockwork Orange, and The Road. It works by determining the quantity and usage of eight different emotions (joy, sadness, anger, disgust, anticipation, surprise, trust, and fear) and two states (positive and negative) throughout the writing.
Videos by VICE
In other words, the program creates an emotional profile based on statistics regarding the presence and frequency of emotion words. Then, based on this data, the system determines the tempo, scale, octave, notes, and sequence of notes before using JFugue (an open source Java API for programming music) to yield an audio file. The music’s key is picked based on the ratio of positive to negative words. The final song chronologically follows the novel, as the music dips and peaks in both tone and tempo based on the text’s narrative arc (we’re not quite sure what happens when a major twist occurs, though).
Though certainly an interesting idea and laudable innovation, there may be some noticable flaws in the program’s logic. For example, Murakami’s Norwegian Wood was transposed into a song (above). The book was originally written in Japanese in 1987 and not published in English until 2000. While the translation might be spot-on (and Murakami himself is bi-lingual), translation is ultimately the “art of failure,” as the English version will never totally replicate the original. Language and culture are too complicated to ever be perfect. So the English version might include emotion words that pertain to sadness, while the Japanese equivalent might be too complicated or cultural different to accurately describe in such terms. The original and the translation as songs would probably sound different. Not to mention, the music genres and instruments picked don’t necessarily reflect the stories’ tones perfectly.
The point here is that the frequency of emotion words used might not be subjective, but emotion is. Not to mention, it can easily be disputed how many sections a story’s narrative has. In general, the interpretation of fiction is too subjective to say “this is what a story sounds like.” Music is too rooted in interpretation, as well. What if I think a story is sad and needs a sad song, but you find it tranquil and peaceful? Do we really trust a program with gaging the emotional value of art?
Also interesting is that several of the novels TransProse morphed into music have already been adapted into films with original scores. Sure, the movies are their own beast and can’t stand as representations for the books, but just try and think about A Clockwork Orange without Beethoven’s 9th popping into your head.
Regardless, this is a fascianting program with interesting applications. I wouldn’t hesitate to even call it a sui generis experiment. The creators said they are even looking into further developments. According to Phys.Org, TransProse will continue to “expolore ways to capture activity in music, [such as] an automatically generated activity lexicon for identifying text portions where characters, fight, dance, or conspire…or where they are relatively passive.” We can only imagine what Hamlet, Infinite Jest, Gravity’s Rainbow, or The Brothers Karamazov would sound like.
h/t Phys.Org
Lead Image via
Related:
This Is What The Audio DNA Of Your Favorite Songs Looks Like
