Faculty Recruiting Support CICS

An Unsupervised Approach to Entity Recognition in Search Queries at TripAdvisor

23 Apr
Tuesday, 04/23/2019 4:00pm to 5:00pm
Computer Science Building, Room 151
Speaker: Rob Hall

Abstract: In order to correctly handle search queries at TripAdvisor it is important to understand which entities (cities, hotels, attractions etc) are being referred to.  For example does the query "london hotel" refer to the London Hotel (which happens to be in Paris) or the more general concept of hotels in London?  In this talk I will describe a noisy channel model which we use to solve part of this problem.  Since we don't have explicitly labeled training data we resort to using an unsupervised approach which makes heavy use of the EM algorithm.  This talk will being with a substantial review of EM and its theoretical underpinnings.

Bio:  Robert Hall holds a joint PhD in ML and Statistics from Carnegie Mellon and a MSc from UMass Amherst. He is currently at TripAdvisor where he works on applications such as ranking and query understanding. His previous position was with the e-commerce company Etsy on ranking and personalized product recommendations.


Faculty Host