Ad-Hoc Georeferencing of Web-Pages Using Street-Name Prefix Trees

A bottleneck of constructing location-based web searches is that most web-pages do not contain any explicit geocoding such as geotags. Alternative solution can be based on ad-hoc georeferencing which relies on street addresses, but the problem is how to extract and validate the address strings from free-form text. We propose a rule-based solution that detects address-based locations using a gazetteer and street-name prefix trees created from the gazetteer. We compare this approach against a method that doesn’t require a gazetteer (a heuristic method that assumes that street-name has a certain structure) and a method that also uses data structures created from the gazetteer in the form of street-name arrays. Experiments using our location based search engine prototype (MOPSI) for Finland and Singapore, show that the proposed prefix-tree solution is twice as fast and 10% more accurate than its rule-based alternative and 10 times faster if an array structure is used when accessing the gazetteer.