Position Validation in Crowdsourced Accessibility Mapping

Abstract

We live in a society in which instant gratification is expected: we demand constantly up-to-date information, which is reflected in our reliance on maps for navigation. Volunteered geographical information (VGI) and geocrowdsourcing make this demand attainable, with popular examples being Waze and OpenStreetMap, where maps are updated quickly by citizen contributors with current base data and features. At George Mason University (in Fairfax, Virginia), the Office of Disability Services releases a traditional paper accessibility map once annually. Owing to its production methods and format, this accessibility map does not capture the transient obstacles that occur frequently throughout campus, rendering it less useful to disabled pedestrians. To fix this dilemma and establish a more useful accessibility system, we have created an application in which contributors report transient obstacles that may impede pedestrian navigation, including sidewalk obstructions, construction detours, and other obstacles that may affect pathway walkability. One of the concerns associated with VGI and geocrowdsourced information is quality assurance, which is imperative when the usage scenarios (including blind, visually impaired, and mobility-impaired navigation) depend on positional accuracy. This study attempts to address the concerns related to the quality assurance of VGI, specifically quality assessment of the positional accuracy of the geocrowdsourced spatial data. We present our quality assessment techniques and novel methods for assessing the consistency of positional characteristics of geocrowdsourced spatial data related to accessibility. These methods rely on moderated positional assessments, geotags extracted from contributed images, and gazetteer-based geoparsing of location descriptions. Finally, we base our methods and approaches on research contributions and best practices from past and current efforts in accessibility mapping.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 12 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.