Sometimes searching for hotel online feels like a blind date: You’ve heard the guy is awesome, but there’s no telling what he’ll be like when you actually meet. A team of researchers from the State University of New York, Stony Brook, might have come up with a way to solve the problem, reports Technology Review’s Neil Savage. Rather than examine individual reviews, they devised a technique that sniffs out the fakes by “pinpointing where the densities of false reviews are for any given hotel,” said Yejin Choi, assistant professor of computer science at the university.
Here’s how: When plotted on a graph, reviews typically produce a pattern that resembles the letter J. That’s because when something is “scored from one to five stars,” said Savage, “it should have a relatively high amount of one-star reviews, fewer twos, threes, and fours, and then a high number of five-star ratings.”
To spot the discrepancy, researchers compared ratings written by frequent reviewers to those from single-time posters to see if the latter were unusually glowing. The larger the gap between negative and positive reviews, the more the J was disrupted—proving the hotel was suspect.
Another tip-off was using too many superlatives, or posting several times within a short time span, often the sign of a marketing campaign.
For consumers, the technology might bring a new age of online booking—that is, if review sites like Yelp and TripAdvisor adopt it. We’d book more confidently knowing the real reviews, and perhaps save some money as well.