130 likes | 271 Views
Supporting Visual Information Extraction from Geospatial Data. Daniele Magazzeni University of Chieti-Pescara, Italy Giuseppe Della Penna Sergio Orefice University of L’Aquila, Italy. Information Extraction .
E N D
Supporting Visual Information Extraction from Geospatial Data Daniele Magazzeni University of Chieti-Pescara, Italy Giuseppe Della Penna Sergio Orefice University of L’Aquila, Italy
Supporting Visual Information Extraction from Geospatial Data Information Extraction Information Extraction aims to extract relevant information from source domain documents
Information extraction approach based on the visual appearance of the information From the low level of code to the higher level of visual features Supporting Visual Information Extraction from Geospatial Data Visual Information Extraction “Whatyouseedrivesyoursearch”
Supporting Visual Information Extraction from Geospatial Data Box Spatial Relations Box relations are effectivefor IE fromdomainssuchas Web pages, electronic documents, butnotsuitablewhengraphicalobjects are representedbycomplexfigures (e.g., in GIS applications)
Supporting Visual Information Extraction from Geospatial Data Image Intersection Problem In the Box Syntactical Model, these two disjoint objects would be intersecting, since their bounding boxes actually intersect
Supporting Visual Information Extraction from Geospatial Data Spatial Relations: Box vs. Contour In the Countour Syntactical Model, the graphical objects have simple polygons as their graphical images, and all the contour points as the corresponding syntactical image Box Contour
Supporting Visual Information Extraction from Geospatial Data Contour Spatial Relations
Supporting Visual Information Extraction from Geospatial Data The SRQ ToolArchitecture The SRQ tool is a full-featured graphical software system to perform VIE on several different source domains
Supporting Visual Information Extraction from Geospatial Data The SRQ ToolQuery Language The SRQL query language has an SQL-like syntax and allows to integrate spatial relations, visual attributes (e.g., color, font) and textual content in the VIE queries
Supporting Visual Information Extraction from Geospatial Data The Malaria Case Study It is well known that the distribution of Malaria is closely related to climatic use SRQ to highlight such a relation by combining and analyzing climate data to find which nations are potentially at risk of malaria Source: FAO, Global Climate Maps, (http://www.fao.org/WAICENT/FAOINFO/SUSTDEV/EIdirect/CLIMATE/EIsp0002.htm)
Supporting Visual Information Extraction from Geospatial Data The Malaria Case Study The SRQ query results are very similar to the actual Malaria diffusion data from World Health Organization! Can we use SRQ to make a forecast of the future diffusion of malaria due to climate changes?
Supporting Visual Information Extraction from Geospatial Data The Malaria Case Study The previous query, applied to a map which indicates the estimated temperature growth in the future (source: UK’s National Weather Service) gives different results: The query results, highlighted in yellow, indicate that the malaria could appear in the future in some areas of the southern Europe as well as in Canada and Australia Interestingly, there have been actually some cases of malaria reported in the last years in Canada and Australia
Supporting Visual Information Extraction from Geospatial Data Thankyou !