Keynote

Prof. Dr. Christopher Jones

Title: Location language, language models and gazetteers

Abstract : Extracting geographical information from text documents is dependent upon determining the locations that the documents refer to. Despite years of research on georeferencing it still remains inaccurate for some applications. Language models of one form or another have emerged to play a key role in categorising entire documents with respect to location as well as with regard to the type of content. Notably some particularly effective methods of georeferencing documents have been hybrid approaches that complement a language model with data from gazetteers. A persistent challenge in georeferencing however is to interpret the meaning of spatial language in locative expressions. That interpretation can be assisted with models of the applicability of spatial expressions, as in spatial templates that can be learnt from examples. This talk reviews examples of applications of different types of language models for georeferencing social media and the locality descriptions in biological records that use complex spatial language. The talk concludes with speculation on the future of gazetteers and the case for their cultural enrichment.

Short Bio: Chris Jones is in the School of Computer Science and Informatics at Cardiff University. His research interests cover aspects of geographical information retrieval relating to georeferencing, information extraction, and interpretation and generation of geo-spatial language. Current applications focus on ecology and on emergency response. Previous research has included map generalisation and multi-scale spatial data access methods, spatio-textual indexing, 3D modelling of buildings and of geological phenomena, and environmental change detection. Early cartographic research on map labelling led to the development of the Maplex map label engine.