Modelling vague places with knowledge from the Web

Place names are often used to describe and to enquire about SPIRIT project D24 3301 Modelling vague places with knowledge from the web IST-2001-35047 v 1.1 Page 2 of 29 geographical information. Geographical place name ontologies, such as gazetteers and geographical thesauri, can help automate geographical information retrieval tasks that employ place names. Most gazetteer resources encoding knowledge of place names are based on official boundaries and provide no data for the many vague places that have no precise boundary but which are commonly used to communicate geographical information. Web search methods are used here to retrieve documents that contain both a target vague place name and the names of associated places, many of which are inside the target place and provide evidence for its spatial extent. Geoparsing and geocoding methods are employed to recognize the presence of the place names and to ground them with coordinates. Given a set of candidate pointreferenced places associated with an imprecise place, two approaches are described to model the extent of the imprecise place. In the first approach techniques based on α-shapes and on Delaunay triangulation are employed to initialise and refine the boundary of the main cluster of candidate points. In the second method the frequencies of occurrence of the associated names are used to generate a density surface model that maps the extent of the target place with varying degrees of confidence. Thresholding the surface at particular density levels enables the creation of a set of possible boundaries. The methods are evaluated by approximating the boundaries of precise places before applying them to delineate imprecise places for which some expert knowledge is available. Note that this report should be read in conjunction with deliverable D25 3302, Generating Imprecise Region Boundaries, which contains details of the technical steps required for the generation of region boundaries. SPIRIT project D24 3301 Modelling vague places with knowledge from the web IST-2001-35047 v 1.1 Page 3 of 29