I’ve been researching the language of online reviews for 6 years now. Prior to that, I had been using TripAdvisor reviews as a useful source of information about hotels whenever I planned to travel.
At some point in late 2008, I began thinking about the potential of TripAdvisor reviews as source of data for discourse analysis. (For those who are unfamiliar with discourse analysis, it’s basically a type of research that explores how people use language – and other symbolic resources – to create specific meanings in a given context. Discourse analysis is the kind of research that I do for my job and, as a professor of Applied Linguistics at a large research university in the Southeast USA, I also get to teach advanced students how to do discourse analysis.)
As I began exploring TripAdvisor’s various site features, my attention was drawn to a section labeled “Rants and Raves,” used by TripAdvisor to showcase the very BEST and the very WORST hotel reviews.
After reading several reviews in both categories, I quickly realized that the “Rants” were a heck of a lot more interesting than the “Raves.” I mean, I am really glad that all those folks decided to share their good experiences too… but they just don’t make for very interesting copy.
In doing some background reading about online reviews, my initial impressions were corroborated by scholars like Ricci & Wietsma (2006) and Sen & Lerman (2007) , who found that readers do tend to pay more attention to negative reviews. (I happen to have a few ideas of my own about WHY this is, and I mention them in Chapter 5 of my forthcoming book.)
Wanting to learn more, I wrote to TripAdvisor several times, asking them “How do you select which reviews to feature in your ‘Rants and Raves’ section?” But they never responded to my inquiries. I am sure their legal department was busy with other, more pressing matters.
Undeterred, I decided to systematically observe the “Rants and Raves” section for several months. I discovered that there were about 5 or 6 different reviews that would cycle through for about one week, and that you could see them all in just few minutes by hitting “refresh.” The next week, another set of 5 or 6 reviews would be featured. And so on.
So that’s how I started my first collection of TripAdvisor reviews. I decided to focus only on the “Rants” since they included lots of gruesome and graphic details: blood stains, insects, inedible food, tales of truly horrible customer service. Plus linguistically, I found that the “Rants” also included a lot of vivid and interesting language: metaphor, hyperbole, extreme case formulations (e.g., This is the worst hotel I have ever stayed in during my 25 years of travel). I downloaded the 5-6 featured “Rants” each week, until I had a dataset of 100 negative reviews. This process took about 6 months.