Improvements in the geocoding process in organizational environments

• Author(s): Correia, Carlos Manuel Oliveia
• Date: 2013
• Persistent ID: http://hdl.handle.net/10362/10864
• Origin: Repositório Institucional da UNL
• Subject(s): Geocoder; Geocoding; GIS; Address analysis
• Document Type: Master thesis
• Language: English
• Advisor(s): Rodrigues, Armanda; Marques, Miguel
• Contributor(s): RUN

Description

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

The current geocoding technologies are only able to handle addresses which fit a general case for the location in consideration. The more edge-case addresses are mostly ignored or wrongly geocoded, leading to imprecision and errors in the results obtained. To try overcoming this problem the current geocoding services accompany their results with confidence values, but the values and scales used vary between services, and are hard to understand by users without knowledge in the area, and, as we discovered, are not truly to be trusted. Novabase aims to make available to organizations a geocoding service which allows the improvement of the quality of the results obtained by mainstream geocoding services such as Google and Bing. The objective is to give quality results in the cases where we can act and, not being able to do, falling back to the results of other geocoding services. We pretend to handle addresses in areas where results are of inferior quality, either because the areas are not fully covered by the services or because those same services are not prepared to handle the address formats which do not match the general case (one example are addresses which are numbered by the use of Lotes). The geocoding is executed in two steps. The first one matches the address with a knowledge base owned by organizations, in which we assume full trust of the quality. If the knowledge base returns a valid result, it is output with maximum confidence. When it fails, we fall back to using the mainstream geocoding services, and use their results for output.