They say you shouldn’t judge a book by its cover, but what about a graph of its emotional trajectory, or character network? A new visualization created by Natalia Bilenko and Asako Miyakawa, two neuroscience PhD students at UC Berkeley, attempts to identify the structural elements of books that might go unnoticed by casual readers.
Bilenko and Miyakawa plotted chapter-by-chapter interactions between characters in The Hobbit, Kafka on the Shore and The Glass Menagerie. Users can see how often Bilbo Baggins occurs in the same passages as Smaug in a circular network diagram. Clicking on Bilbo’s name highlights the peaks and valleys of his emotions throughout the entire novel in a bar chart seen below.
Some scholars view this kind of narrative visualization as a way to democratize the literary canon. By setting computer parsers loose on collections of millions of texts, we can track how our world has been reflected in writings of the time. The recent mash of data and literature has been called “culturomics,” “distant reading,” and “macro-analysis,” variously.
But while a chart showing every time Gandalf calls the Great Eagles is neat–hell, we’d read an entire book about those birds–does it add anything to the reading experience? Bilenko and Miyakawa argue yes, provided that they can automate the visualization and apply it to any book. Using reading history as input, the two imagine users identifying preferences they hadn’t even considered. “People could get a subjective opinion from book reviews, but also have access to an emotional trajectory that’s a little more objective and data-driven,” Miyakawa says.
Amazon already offers selections based on your purchase history, and sites like Pandora and Netflix have used structural data to create more accurate genre selections for users. Could it be long before Amazon greets us with personalized recommendations for “Dickensian Ensemble Pieces with Happy Endings”?