As reported last week, TripAdvisor is the most-visited pre-transaction travel website in the world - meaning hundreds of millions of visitors each month come to the site looking for travel recommendations.
That’s a lot of data for the platform to host - and for users to sift through - which is why TripAdvisor has turned to artificial intelligence to help make sense of its review content.
Subscribe to our newsletter below
Jeff Chow, TripAdvisor vice president of product, consumer experience, says the technology assists in personalizing recommendations on the site - that is, sorting the platform’s user-generated reviews and delivering them to the right visitor at the right time, depending on where he or she is at in the trip-planning process.
“Not every millennial wants to live like a local; not every family wants to live like a tourist. How do we match interest with the right community voices that we have?” Chow says.
“Frankly, [with AI] you can make better and faster decisions than before when you had to read every single review and filter every single review.”
How does it work?
Chow says machine learning - which powers AI - is used to “derive meaning” out of TripAdvisor’s massive data sets, and AI is then “married to other techniques such as UX, interface design to make it more humanized.”
The big lever that we’re always trying to fulfill is FOMO.
Jeff Chow - TripAdvisor
A process called collaborative filtering matches users with specific interests to reviews that include similar content - in essence looking for patterns to surface relevant reviews to each traveler.
Sentiment analysis helps determine where users are in the planning cycle as well as what types of reviews they’re looking for: Someone narrowing down a selection of 10 hotels, for example, likely won’t need to read full reviews for each property, however a traveler ready to make a purchase might want more information.
Examining this data “ultimately helps us understand the traveler journey,” Chow says. “Understanding when to surface the meat and when to surface the soft signals is a huge challenge and what we rely on AI and machine learning for.”
How does it grow?
Chow says the difference between travelers visiting TripAdvisor compared to online travel agencies is intent. TripAdvisor users might be researching locations or browsing in-destination activities, whereas those on an OTA website have intentions to book.
“We want to understand where the traveler is in every cycle so we can deliver the best info they can get, starting with destination selection all the way through lodging decision,” he says. “The big lever that we’re always trying to fulfill is FOMO.”
He says the TripAdvisor machine learning team has been building up a body of work for the past three years to help analyze and understand review content, starting with basic functionality around things like tagging and filters.
We’re not trying to replace our amazing communities; we’re trying to match travelers with the community members that give you the best recommendation.
Jeff Chow - TripAdvisor
In the past year, “we took really seriously the aspect of personalization, taking the wisdom of the [review] crowd and all of that info and marrying it with what the traveler is actually seeking,” he says.
“Our whole thesis is the technology is a humanizing moment. We’re not trying to replace our amazing communities; we’re trying to match travelers with the community members that give you the best recommendation.”
Chow says the technology is “early days” in general, but the future is integration with digital assistants and voice technology to give more confident recommendations.
Voice will become part of a “multimodal” solution to develop a full holistic experience, he says, where travelers can seamlessly continue trip planning across devices.
Although that frictionless ideal will take time, AI is still fundamental in how TripAdvisor manages data and increases personalization across its reviews platform.
“Our big, hairy, audacious goal is to be your travel assistant,” Chow says. In order to scale, “we need machines to help us do that.”