Neighborhood Self-Identity and Point of Interest Identification on Airbnb
Date
Authors
Thomas, Peter
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Room-sharing marketplace Airbnb is disrupting the short-term rental market and leisure travel industry by providing a platform to connect accommodation producers (Hosts) and consumers (Guests). Airbnb is growing rapidly and has more than 3 million Listings worldwide. Airbnb Listings are a rich and under-researched corpus of Volunteered Geographic Information (VGI). A Listing’s neighborhood description, written by the Host, contains a wealth of information on local attractions such as parks, restaurants, nearby landmarks, and neighborhoods. This thesis uses the neighborhood descriptions in geolocated Airbnb Listings to delineate neighborhood boundaries and discover and geolocate unique Points of Interest (POIs) in New York City. This study constructs neighborhood maps through DBSCAN convex hull creation and hex assignment. Results show that the context of a neighborhood name changes based on how early it occurs in the neighborhood overview field: the first sentence is the Listing’s location, and subsequent sentences are references to other nearby neighborhoods. Network analysis demonstrates that Listings reference nearby neighborhoods frequently and distant neighborhoods infrequently. The DBSCAN clustering algorithm is applied to effectively identify which frequent ngrams are highly spatially clustered and likely to represent a unique POI. This work is a novel application of crowdsourced neighborhood and POI identification techniques to a new VGI dataset.
Description
Keywords
Volunteered Geographic Information, Airbnb, Neighborhood, Point of interest