Automatic discovery of global and local equivalence relationships in labeled geo-spatial data.

We propose a novel algorithmic framework to automatically detect which labels refer to the same concept in labeled spatial data. People often use different words and synonyms when referring to the same concept or location. Furthermore these words and their usage vary across culture, language, and place. Our method analyzes the patterns in the spatial distribution of labels to discover equivalence relationships. We evaluate our proposed technique on a large collection of geo-referenced Flickr photos using a semi-automatically constructed ground truth from an existing ontology. Our approach is able to classify equivalent tags with a high accuracy (AUC of 0.85), as well as providing the geographic extent where the relationship holds.