Deriving Locational Reference through Implicit Information Retrieval

GIScience 2016 Short Paper Proceedings Deriving Locational Reference through Implicit Information Retrieval T. Hervey, W. Kuhn Department of Geography and Center for Spatial Studies, University of California, Santa Barbara, CA 93106 Email: {thomas.hervey; kuhn} @geog.ucsb.edu Abstract The often fragmented process of online spatial data retrieval remains a barrier to domain scientists interested in spatial analysis. Although there is a wealth of hidden spatial information online, scientists without prior experience querying web APIs (Application Programming Interface) or scraping web documents cannot extract this potentially valuable implicit information across a growing number of sources. In an attempt to broaden the spectrum of exploitable spatial data sources, this paper proposes an extensible, locational reference deriving model that shifts extraction and encoding logic from the user to a preprocessing mediation layer. To implement this, we develop a user interface that: collects data through web APIs and scrapers, determines locational reference as geometries, and re-encodes the data as explicit spatial information, usable with spatial analysis tools, such as those in R or ArcGIS. 1. Introduction GIS advancements have produced a growingly complex general-purpose toolbox rather than functionality tailored to domain-specific questions. Frequently, domain scientists including Green (2015: 717) highlight the salient lagging data and tool limitations associated with GIS. As Kuhn (2012: 2267) notes, it is essential to rethink the fundamentals of spatial information while promoting clarity that cuts across technical boundaries and broadens spatial literacy for non- experts. Contributing to the work by Kuhn and Ballatore (2015: 219) and Vahedi et al. (2016) to design an intuitive GIS language for question-driven spatial studies, we focus on bridging the gaps between data discovery and spatial analysis tools by broadening the spectrum of exploitable spatial data sources. Compared to the vast amount of implicit spatial data (hidden location attributes often in the form of metadata, auxiliary place names, and geotagged attributes (Heinzle and Sester 2003: 335)), there remains a relatively limited quantity of online explicit spatial data (georeferenced geometry-based features (Brisaboa et al. 2011: 358)). When available, explicit data are typically served from a limited number of administrative portals or require intensive energy and time from a user searching, exporting, encoding and cleaning before being usable (Munson 2013: 65). These preprocessing requirements limit the feasibility of question-driven spatial analysis (Vahedi et al. 2016) and force domain scientists to base their studies on data availability. It is clear that implicit spatial data is an attractive alternative. However, as the numerous research challenges associated with GIR (geographic information retrieval) suggest (Jones and Purves 2008: 219), current methods do not provide adequate solutions to navigate, gather or utilize the mass of heterogeneous implicit spatial data spread across the array of online repositories. Custom-constructed web API requests and scrapers can help retrieve and process this unpublished information. Yet, without technical expertise to build new or use existing