We are swimming in an ocean of online reviews. However, when you look at individual sites, whether it be Travelocity, Orbitz or the like, you have more of a lake of reviews, a much smaller pool. There’s a lack of content.
Then you go to the larger sites, like TripAdvisor or Booking.com, where the content is more robust, but who has time to read it all? Yes, all of the feedback is valuable, but you might as well take off a year and conduct a master thesis on the destination if you are going to go through all of the reviews.
That’s when a context problem arises; there just aren’t enough hours in the day to sort through it all or a good way to filter it. I want the pros and cons of a hotel based on travelers’ past experiences. And I would venture to bet I am not alone.
For the statistic freaks out there, yes size does matter! We have evaluated reviews from different sources (OTAs versus review sites) and different languages and the numbers tell us the more reviews we aggregate, independent of source or language, the better the “objective” end result. It’s the law of large numbers.
So let’s assume for a second we have solved the content problem (for the most part TripAdvisor has). When you are the king (content), you should start looking for the queen (context in this case). Context means understanding. Understanding means in the world of big data “structuring the content”. Once your database is full of beautiful structured content, you can use it in various ways.
One big shortcoming in the travel industry is still the “search problem”. There is no website out there that allows you to search for the “best family hotel in Rome” based on user-generated reviews. Sure, I can search that general term and I’ll come up with a long list of properties that have declared themselves the best in the category (or simply were savvy with links and key words).
I want to know what the actual guests – people who stayed at these hotels with kids – thought about these accommodation when they were traveling with their families. And I want all the hotels ranked based on all the thousands of positive or negative comments. Or why can’t I search for “clean rooms in Bangkok”? I know that cleanliness is one of the top-mentioned comments in reviews so it must be important, yet still I cannot search for it. Bummer!
By aggregating all of the content, and then analyzing all of the review data (i.e. dissecting the context of what is being said), you come up with the pros and cons that can be used for search and filtering purposes. Categories such as “clean rooms,” “business travel hotels,” “family friendly,” “free and strong WiFi” are easily dictated by the experiences of past guests who use these terms to describe their experience.
I trust this much more than a business travel hotel that claims to be family friendly to fill weekend need periods. The more and more people who say this given hotel is family friendly, the more believable it is; more so than a hotel saying it themselves.
Once I quickly nail it down to the top three hotels in my search, I need a different presentation layer for the hundreds or thousands of reviews per property. Because the score itself is not enough for me and, as mentioned earlier, I just have no time to read all of this.
But this doesn’t mean I don’t want to know what they all have had to say. Why not let computers read this for the users? READ ONE AND DONE! Yes, this is possible with sophisticated semantic analysis. Let’s give the users a high-level summary of these reviews, like the cliff note version of a lengthy book, so that we can make better educated decisions in less time based on ALL the user-written reviews available.
Keep an eye on these sites; filtering and semantic analysis with context-based search features are going to come into play. At the end of the day, content is king and context is queen. And what’s a kingdom without a queen?
Note: Benjamin Jost will be speaking at the WIT Conference, Oct 21-23 in Singapore.