Dartmouth Researcher Mines Amazon.com To Measure Literary Tastes

Gronas, an Assistant Professor of Russian Language and Literature, is interested in literary tastes. He wants to know why people read certain books, what drives those reading decisions, and what lies behind readers' reactions. Sociological surveys are fine, he says, but the answers are shaped by the questions. With online book reviews, like those at Amazon.com, he can begin to get a quantitative measure of taste (from the number of stars assigned by readers to a book) along with a qualitative assessment (from the personal commentary provided by readers).

This research was recently presented at the American Association of Teachers of Slavic and Eastern European Languages conference in Boston last month.

"Amazon.com book reviews are not based on literary theory," he says. "They are written by everyday readers, not scholars, who bring a new perspective to the topic of taste. Since online reviews are voluntary, they offer honest opinions that aren't prompted by specific questions."

With help from a colleague, Maksim Yegorov, a graduate student at the University of California at Berkeley, who wrote a computer program to collect the information, Gronas first studies the ratings that are assigned to books by Amazon.com customers. The "classics" usually get consistently high grades, in the four or five star range, albeit from fewer reviewers. Classics also have fewer detractors, or "taste dissenters," than other categories of books. Books with a large fan base, such as the books in the Harry Potter series, usually attract thousands of online reviewers, who also dispense high grades. The grade analysis produces predictable and regular patterns of taste.

"I can plot the grades of Tolstoy's Anna Karenina and Crime and Punishment and see similarities, and do the same for the Harry Potter books," says Gronas. "Similar books have similar audiences that react in similar ways, which is not surprising." He says it was astonishing, though, how precisely the numbers lined up.

On Amazon.com, there are also books that have what Gronas calls a high "index of controversiality." It's a measure that indicates varying opinions. The overall grade average of the books may be similar, but because the grades are more spread out, the index most often suggests a topic where opinions diverge and controversy arises. Take, for example, Rush Limbaugh Is a Big Fat Idiot by Al Franken and The Way Things Ought to Be by Rush Limbaugh.

"Each of these books reflects opposing views, and probably they each attract a different audience, so you might conclude that they have little in common. Their grade profile, however, is remarkably similar. I suspect that books that have a high 'index of controversiality' are more likely to sell better."

The second part of Gronas' work involves digging deeper into the reviews to understand the more qualitative or subjective elements of literary taste. He says that the reviews usually contain some language that implies a visceral or emotional effect, say if a book makes a reader happy or depressed. Readers often comment about how long a book took to get through or whether a book was boring. There are also reviews with more intellectual commentary, with more complex phrases and richer vocabulary. Some of the reviews, according to Gronas, have a blog-like quality to them with short sentences, quick exclamations as well as slang. Some are witty and creative.

"It's generally believed that people with emotional tastes are separate from people with intellectual tastes, that these two characteristics can't reside in the same person," he says. "In the Amazon.com book reviews, I often saw both of these elements of taste illustrated in the same commentary of a reviewer. One person can exhibit a hybrid of tastes."

Gronas says that the online review community is a virtually untapped wealth of information that provides insight into what shapes opinions and cultural preferences. Individual choices and judgments have been studied by economists and social scientists. Amazon.com adds another dimension to this field by providing a new pool of data to examine.

"I am introducing a palpable, probabilistic approach to literary criticism. That's what makes it fun."