Intelligent Information Retrieval for Maps

Analysis of spoken and search engine queries showthat people often look for maps of acertain location that are at a certain time or show asubject theme. Their search may behindered by the fact that maps in both physical anddigital collections are often organizedby region, and maps in non-geographical publicationsare essentially invisible to searchbecause they are rarely indexed. This research offersa way to resolve the problem bywriting algorithms that automatically classify mapsby time and theme as well as region,demonstrating these techniques in a web-basedprototype system. The classification is"intelligent" because it is aided by knowledgeresources. These automatic classificationwere evaluated by comparing system-classified maps tomanual classification of the samemaps, with results showing automatic correspondencewith the manual at 75% for region,69% for time, and up to 93% for theme (when plausibleclassification categories arecounted). The discussion will appeal to those whostudy information retrieval as well asto those who just like maps.