TripAdvisor & Text Mining

  1 m, 31 s

Here’s the second video taken from our LUG conference.

This video showcases John Kelley, from TripAdvisor, discussing how they use text analytics on the reviews section of their website, in order to determine how their users interact with the search functions. TripAdvisor, being one of the most popular sites out there, tends to contain a ton of reviews per location, which can be really problematic for users who want precise, quick information. Also, when a user is presented with too much data, the question is no longer “what do others find important”, but rather “what’s important to me?”, 

John demonstrates how, through their usability testing, they determine that phrases work better than concepts, and that needlessly complex options tend to distract users rather than help them. As such, they built a concept generator which summarized reviews, generating 2-3 word concepts from each one, and displayed 3-10 concepts per property. They then used Lexalytics’ Salience text analytics engine to add sentiment analysis to these concepts, which helps users sift through the positive and the negative, allowing them to make an easier choice.

It’s fascinating to note that by simplifying their site, users became much happier. Apparently, 3 flavors of ice cream work better than 30 (but don’t tell that to Baskin Robbins).

John Kelley is a Sr. Product Manager for Consumer Products at TripAdvisor.com. TripAdvisor® is the world’s largest travel site with over 50MM unique visitors per month, and 60MM+ user contributed reviews. John and his team focus on optimizing the selection process for accommodations, restaurants, and attractions by creating the best experience to delight the consumer. Prior to joining TripAdvisor, John held senior product positions at Zoom Information, Monster Worldwide, Akamai Technologies, and NewsEdge Corporation. John is a graduate of Bentley University and the University of Massachusetts, Amherst. 

Categories: Events, Lexalytics User Group, Partners, Presentations, Sentiment Analysis, Text Analytics