Historical Gazetteer System Integration: CHGIS, Regnum Francorum, and GeoNames

Integration of digital gazetteers, involving the disambiguation of unique places and conflation of duplicates or variant placeneames, has been the focus of many theoretical papers in recent years. The challenge of mapping between historical instances of placenames is also an ongoing concern for several important projects dealing with ancient placenames. Here the matching of historical placenames from two unrelated datasets to gazetteer web services is undertaken using a simple geospatial and geonomial algorithm. The quantitative results of the matching trials are considered, problems in dealing with vernacular scripts considered, and practical implications for integrating historical gazetteers discussed. 1. Existing Gazetteer Web Services and Digital Historical Gazetteers The importance of online gazetteer services as authorities for geocoding placenames, or retrieval of placenames based on queries containing real world coordinates (reverse geocoding) has already been demonstrated. For example, the expanding interconnections of LinkedOpenData 1 on the semantic web have consistently placed GeoNames 2 at or near the center of the semantic web's social graph. The centrality of GeoNames is largely due to three factors: first, it is a free and open API; second, the API is simple and easy-to0-use; and third, it is currently the only global geographic resource with stable URIs. Traffic for the GeoNames web service has topped 20 million requests per day, and since half of these are from smart phones, 3 it is clear that geographic information retrieval [GIR] is being built into many new location-based applications for hand-held devices. Another major GIR web service is provided through the GoogleMaps Geocoding API,4 which provides free geocoding and reverse geocoding web services. But the GoogleMaps web service, unlike GeoNames, provides no standard URI or unique identifier with their query results, which explains why there is no GoogleMaps presence on the OpenLinkedData cloud. Even so, the general explosion of webmaps and geocoding applications based on the GoogleMaps is clear to be seen. In 2010, Google declared that more than 350,000 websites were using the service, and that: "Google Maps API has established itself as the most popular Google API and the most deployed service-based API on the web." 5 Clearly, the demand for accurate, automated GIR has become an essential part of the Internet experience for a rapidly growing audience. The emergence of these robust gazetteer web services -GeoNames, GoogleMaps Geocoding API, Yahoo Placemaker 6 -provides an interesting testbed for GIR research. They have clearly outstripped their predecessor, the Alexandria Digital Library [ADL] gazetteer content standard and protocol, in terms of performance.7 And while ADL established the basic principles of digital gazetteers, 8 the new breed of gazetteer web services simply appear as operational APIs, with technical documentation on query and response parameters but no theoretical underpinnings at all. Therefore it is quite interesting to see the ways in which GIR research is taking advantage of these gazetteer web services, by trying out new methodologies for integrating digital gazetteers, as well as exploring new theoretical aspects of GIR on the semantic web. 9 One prospect for integration of digital gazetteers, is to augment the existing gazetteers with temporal attributes, turning them into spatio-temporal gazetteers, and enabling Geo-Temporal Information Retrieval [GTIR]. An obvious place to begin with this task, would be to establish links from the dated placename attestations in existing historical gazetteers to the undated placenames