Learning Location Naming from User Check-In Histories

Many innovative location-based services have been established in order to facilitate users’ everyday lives. Usually, these services cannot obtain location names automatically from users’ GPS coordinates to claim their current locations. In this paper, we propose a novel location naming approach, which can provide concrete and meaningful location names to users based on their current location, time and check-in histories. In particular, when users input a GPS point, they will receive a ranked list of Points of Interest which shows the most possible semantic names for that location. In our approach, we draw an analogy between the location naming problem and the location-based search problem. We proposed a local search framework to integrate different kinds of popularity factors and personal preferences. After identifying important features by feature selection, we apply learning-to-rank technique to weight them and build our system based on 31811 check-in records from 545 users. By evaluating on this dataset, our approach is shown to be effective in automatically naming users’ locations. 64.5% of test queries can return the intended location names within the top 5 results.

In  ACM GIS 2011

Publisher  ACM

Details

TypeInproceedings
> Publications > Learning Location Naming from User Check-In Histories