Budget Travel's Blog

The latest travel news, vacation tips & advice, travel apps, and more

aug 15 2011

A computer algorithm that spots fake user reviews? Yup!

How many times have you researched hotels on TripAdvisor (or any user review forum, for that matter), and found yourself wondering how accurate the descriptions were? We all know to discredit excessively positive or negative reviews (which were likely written by someone with an agenda), but what if there was a reliable way that sites could flag suspicious write-ups?

Cornell students and professors have come to the rescue. As reported by Travel Weekly, the tech geniuses at the Ivy League college have created a software program that can identify fraudulent critiques with 90% accuracy (human judges, on the other hand, were only accurate 50% of the time).

To put their algorithm—which looks for keywords that are more likely to be embedded in fake reviews—to the test, students hired writers to create 400 false reviews for TripAdvisor's 20 most popular hotels in Chicago. These reviews were then interspersed with 400 actual reviews and subjected to the software program for analysis. Human test subjects formed the control group. As indicated above, the results suggest that no matter how hard we try, we humans aren't good judges of honesty online. (The students published their findings in a report entitled "Finding Deceptive Opinion Spam by Any Stretch of the Imagination.")

The issue of fake reviews is not a new one. It's been on travelers' minds as far back as 2007, when Nancy Keates called attention to the problem in a telling article for the Wall Street Journal. Two years later TripAdvisor started posting disclaimers warning customers to watch out for fake reviews (the subject inspired a lengthy debate on this site about the trustworthiness of user-generated review sites).

The debate, however, seems more timely than ever when you consider that TripAdvisor—already the biggest travel review site on the web—has grown its traffic by 35% since 2008.* Whether or not the site jumps to use the program created by Cornell remains to be seen, but I for one am glad to see that there are folks out there who are actively searching for a solution.

What about you? Do you worry about being misled by fake information on user review sites, or do you think you're a pretty good judge of accuracy? Do you have any tips on how to avoid being taken by misinformation?

*according to PhoCusWright research quoted by Travel Weekly.

SEE MORE FROM BUDGET TRAVEL:

How is Airbnb Dealing With the Robbery That Rocked the Vacation Rental World?

Open Secret Websites for Booking Hotels

8 Ways to Save on Hotel Rooms

Get Inspired with more from BudgetTravel.com


Note:This story was accurate when it was published. Please be sure to confirm all rates and details directly with the companies in question before planning your trip.

Video


Loading Comments...

Our newsletter delivers vacation inspiration straight to your inbox.

Check Prices