Given a set of photographs labeled with both geographic location and time, this paper explains how to cluster the photographs, and then name the clusters. The goal is to select short and recognizable geographic names, and to do this, the presented algorithm uses the count of Google retrievals for city, state, and park names to select the most suitable ones. Short experiments verified that the algorithm chooses good names.
This paper is straightforward, easy to read, and demonstrates a clear improvement over less careful ways of picking geographical labels.