Fuzzy representation of vague spatial descriptions in real estate advertisements

Abstract

Geocoding a spatial description is challenging since vernacular place names and vague spatial expressions give uncertainty and ambiguity to the description. Usually, digital gazetteers are used to match geospatial objects to their boundaries. However, gazetteers do not contain all places. Therefore, a number of studies have proposed to enrich gazetteers by estimating and representing the vernacular places. Nevertheless, only a few approaches have taken into account vague spatial expressions such as "nearby", and have represented geospatial objects as sharp boundaries. In this work, we present an automatic workflow to retrieve a location approximation of vague spatial description. We propose a model to estimate a fuzzy representation of each mentioned geospatial information and spatial expressions. Then, we perform information fusion to find a location approximation of a property. Lastly, we demonstrate our proposed method by applying it to the case of French Real Estate advertisements with two real-world datasets in Nice and Paris. Real Estate advertisements allow us to deal with uncertain geospatial objects since avague and exaggerated property location's description is usually provided. Our results show that our proposed method is promising and able to correctly approximate a location from uncertain spatial descriptions.