Authors: Alexander Savelyev*, Texas State University
Topics: Cartography, Geographic Information Science and Systems
Keywords: textual geography, cartography, typography
Session Type: Virtual Paper
Start / End Time: 9:35 AM / 10:50 AM
Room: Virtual 16
Presentation File: No File Uploaded
Exploratory analysis of spatial trends and patterns in geographic text datasets is hypothesized to have potential for a wide range of applications. A common component of exploratory text analysis is automated keyword extraction, and numerous attempts were made at adapting it to the tasks of exploratory spatial analysis. Unlike traditional keyword extraction algorithms, however, it is unclear how to evaluate the quality of the results produced by the spatial keyword extraction algorithms. We propose a user study design that compares the results of a prototypical spatial keyword extraction algorithm (quadtree-based tf-idf) to a multi-scale collection of spatial keywords produced by a group of human subjects. Throughout the study, we approximate the hypothetical scenario of producing high-resolution multi-scale maps of surnames for demographic studies in geography, using 6 million spatial surname records obtained from the Ohio electoral roll dataset. The results of the study can be interpreted as a step towards building a empirically-derived ground truth dataset (along with the methodology for producing one) for a robust and automated evaluation of spatial keyword extraction algorithms.