Location Estimation Using Crowdsourced Spatial Relations

The "crowd" has become a very important geospatial data provider. Specifically, nonexpert users have been providing a wealth of quantitative geospatial data (e.g., geotagged tweets or photos, online). With spatial reasoning being a basic form of human cognition, textual narratives expressing user travel experiences (e.g., travel blogs) would provide an even bigger source of geospatial data. Narratives typically contain qualitative geospatial data in the form of objects and spatial relations (e.g., "St. John's church is to the North of the Acropolis museum." The scope of this work is (i) to extract these spatial relations from textual narratives, (ii) to quantify (model) them, and (iii) to reason about object locations based only on the quantified spatial relations. We use information extraction methods to identify toponyms and spatial relations, and we formulate a quantitative approach based on distance and orientation features to represent the latter. Probability density functions (PDFs) for spatial relations are determined by means of a greedy expectation maximization (EM)-based algorithm. These PDFs are then used to estimate unknown object locations. Experiments using a text corpus harvested from travel blog sites establish the considerable location estimation accuracy of the proposed approach on synthetic and real-world scenarios.