Search papers, labs, and topics across Lattice.
This paper introduces a novel approach to location retrieval in Airbnb search by rearchitecting the system to retrieve from a subset of highly bookable, high-precision rectangular map cells, dividing the world into 25 million uniform cells. This method addresses the challenge of balancing a diverse supply of homes with varied guest expectations in a two-sided marketplace. The rearchitected search system improves retrieval efficiency by focusing on the most relevant geographic areas before applying resource-intensive ranking models.
Airbnb's new search architecture carves the world into 25 million cells and uses extreme classification to predict which ones a user is most likely to book from, boosting retrieval precision.
Airbnb search must balance a worldwide, highly varied supply of homes with guests whose location, amenity, style, and price expectations differ widely. Meeting those expectations hinges on an efficient retrieval stage that surfaces only the listings a guest might realistically book, before resource intensive ranking models are applied to determine the best results. Unlike many recommendation engines, our system faces a distinctive challenge, location retrieval, that sits upstream of ranking and determines which geographic areas are queried in order to filter inventory to a candidate set. The preexisting approach employs a deep bayesian bandit based system to predict a rectangular retrieval bounds area that can be used for filtering. The purpose of this paper is to demonstrate the methodology, challenges, and impact of rearchitecting search to retrieve from the subset of most bookable high precision rectangular map cells defined by dividing the world into 25M uniform cells.
