System Demo: A geo-coding service encompassing a geo-parsing tool and integrated digital gazetteer service

A geo-coding service encompassing a geo-parsing tool and integrated digital gazetteer service Ian Densham Edinburgh University Data Library Main Library Building George Square Edinburgh EH8 9LJ Scotland, UK i.densham@ed.ac.uk James Reid Edinburgh University Data Library Main Library Building George Square Edinburgh EH8 9LJ Scotland, UK james.reid@inf.ed.ac.uk Abstract We describe a basic Geo-coding service en- compassing a geo-parsing tool and integrated digital gazetteer service. The development of a geo-parser comes from the need to explicitly georeference large resource collections such as the Statistical Accounts of Scotland which cur- rently only contain implicit georeferences in the form of placennames thus making such col- lections inherently geographically searchable. 1 Introduction The project is being undertaken by the Edinburgh Univer- sity Data Library (Edina) as a part of the larger geoXwalk (www.geoXwalk.ac.uk) project which aims to develop a protocol based (ADL, OGC and Z39.50) UK digital gazetteer service. The geo-parser uses the geoXwalk server as the name authority for identified placename can- didates. 2 The geo-coding process In its current implementation, the service consists of two main components, the geo-parser and the geoXwalk gazetteer, with a generic demonstrator interface. The term geo-parsing refers to the identification of place- names in a document/resource, where geo-coding refers to the tagging of the candidate and consequently the re- source with a geographic footprint. Figure 1 shows the basic geo-coding flowline. A resource is submitted to the geo-parser, which iden- tifies a series of potential placenames. Each placename is displayed along with the number of occurrences in the text, and the number of matching gazetteer candidates. For each placename, a link to the gazetteer records is displayed and a highlight option is available for identi- fication in the original text which is displayed beneath Figure 1: The geo-coding process the table. Various sorting functions are also available for records of the table. County and feature type are the default attributes for disambiguation, although more are available through the geoXwalk feature specification. Currently multiple gazetteer entries can be attached to a single placename, enabling output of different instances of the same name in the text. Geo-coding output is avail- able in an application specific xml schema, csv, or html, and contains parser and editor matadata. Outputted pla- cenames can be viewed on a map. Clearly the degree of human interaction is high duri! ng the review stage, with the process currently limited to individual resources. As geo-parser development continues, user interaction at this stage of the process will become less, although the po- tential for ‘post process’ queries will rise, as the parser is more closely integrated with the geoXwalk database. As geo-parser development progresses the interface will need to accommodate a more flexible approach to the geo-coding process, as clearly interface requirements are determined by users with associated collections of spe- cific document types, and output requirements. A range of functionality is required at various levels between a fully automated batch processing mode and a more inter- active analytical approach to individual documents. Fur- ther investigation is required …

Content truncated to 500 words