Geospatial database generation from digital newspapers: use case for risk and disaster domains.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies.

The generation of geospatial databases is expensive in terms of time and money. Many geospatial users still lack spatial data. Geographic Information Extraction and Retrieval systems can alleviate this problem. This work proposes a method to populate spatial databases automatically from the Web. It applies the approach to the risk and disaster domain taking digital newspapers as a data source. News stories on digital newspapers contain rich thematic information that can be attached to places. The use case of automating spatial database generation is applied to Mexico using placenames. In Mexico, small and medium disasters occur most years. The facts about these are frequently mentioned in newspapers but rarely stored as records in national databases. Therefore, it is difficult to estimate human and material losses of those events. This work present two ways to extract information from digital news using natural languages techniques for distilling the text, and the national gazetteer codes to achieve placename-attribute disambiguation. Two outputs are presented; a general one that exposes highly relevant news, and another that attaches attributes of interest to placenames. The later achieved a 75% rate of thematic relevance under qualitative analysis.