GREASE-II researches information access methods to large collections of documents and objects having geographically rich text and meta-data, with emphasis on the web. GREASE-II the follow-up of GREASE (POSI/SRI/47071/2002) breaks the assumption taken at the GREASE project that the geographic scope of a document could be abstracted to a single geographic concept or shape, which could be obtained from meta-data or by applying information extraction techniques to the text and/or graph algorithms to its hyperlinks. The geographic content of a document is now characterized by a summary made of a set of geographic tags, similar to human added tags in web2.0 applications. Some of the tags will map directly into ontologic geographic concepts, while others may not even be geo-referenceable. We hypothesize that these summaries, which may also be derived from the contents using information extraction methods, fuzzy reasoning and graph algorithms, provide a much richer description of the contents. Using richer abstracts to represent the geographic area of interest of documents will significantly improve geographic searches, enabling the surpassing of the hurdles imposed by the simpler scopes-based approach explored on GREASE.
LASIGE is supported by FCT, project UID/CEC/00408/2019