By Deborah Yaffe, Jun 25 2018 01:00PM
Back in high school, I did reasonably well in math. I may have trouble balancing my checkbook these days, but I promise I’m not one of those math-o-phobes who thinks numbers are the work of the devil.
Still, at the risk of sounding like a STEM Philistine, I can’t help but scratch my head at what seems to me to be the sheer pointlessness of a math-enhanced literary analysis spotlighted last month by the BBC.
Many years ago, apparently, the novelist Kurt Vonnegut concocted an anthropological theory that he summed up thusly: “Stories have shapes which can be drawn on graph paper.” More recently, math-y folk have used a statistical text-mining technique called “sentiment analysis” – basically, a way of coding the emotional content and associations of vocabulary words – to identify six archetypal story arcs characterized by the rising and falling patterns of the protagonist’s fortunes.
Summarizing academic work that’s been around for several years, the BBC provides us with the story-arc map that sentiment analysis creates for six famous works of literature -- including Pride and Prejudice, used to illustrate the “Cinderella” plot type (rise, fall, rise). “While not a perfect tool – it looks at words in isolation, ignoring context – [sentiment analysis] can be surprisingly insightful when applied to larger chunks of text,” the BBC opines, citing as an example of such insight a 2016 blog post on P&P by Janeite data scientist Julia Silge.
Sentiment analysis, as far as I can glean from Google, is a tool whose main purpose is commercial: It allows businesses to quickly sift online reviews, comments, and social media posts in order to pick up trends in consumer reactions and resolve problems before they metastasize. In other words, it’s designed for people who want to know the essence of what’s being written without actually having to read anything.
By contrast, literature – pace SparkNotes – is not about skimming to get the gist. It’s about reading carefully to get the nuance. From an anthropological point of view, it may be interesting and worthwhile to categorize the pattern in our storytelling – or the six patterns. But whatever sentiment analysis may have to teach us about the broad outlines of stories, it doesn’t have much to say about individual works of art, or about the subtleties that make great writers worth reading.
Can Austen’s novel be slotted into a plot archetype? I wouldn’t be surprised – human storytelling is an ancient art, after all – but filing P&P under Plot Type #5 doesn’t tell us much of anything about what makes it better than, say, all the P&P fanfic that mimics its structure.
Indeed, looking at Silge’s P&P plot graph suggests the shortcomings of the sentiment-analysis approach as applied to a specific book. For instance, Silge finds that the Netherfield Ball is the high point of positive sentiment in the novel, and it’s true that much excitement and anticipation attend that scene. But it’s also the locus of much disappointment and mortification for Elizabeth Bennet: Wickham mysteriously absents himself, Darcy surprises her into an unwanted dance and a tense conversation, and her mother, sister, and cousin commit public faux pas.
How does placing the ball at the top end of a sentiment scale help us unpack the complicated spectrum of feeling across which the scene in the novel actually ranges? Not at all, as far as I can tell. Arguably, Silge's graph, at least if it's used to analyze a specific book rather than to identify broad patterns across books, is not just pointless but actively misleading. And at the points where the graph matches any casual reader's impression of the book -- for example, in identifying Lydia's elopement as the story's emotional low point -- it's superfluous.
So file me under Unconvinced. But, then, I was a humanities major.