A survey of open accessibility data
Chaohai Ding, Mike Wald, Gary Wills · 2014 · Proceedings of the 11th Web for All Conference (W4A) · doi:10.1145/2596695.2596708
Summary
This paper surveys the landscape of open accessibility data in the UK, focusing on geographic and location-based datasets that could help people with disabilities navigate physical spaces and plan travel. The authors examine five data sources: Wheelmap (a crowdsourced map of wheelchair-accessible places built on OpenStreetMap), Factual (a commercial location platform with restaurant accessibility data), Transport for London's step-free access guide for tube stations, UK National Rail station accessibility data, and AccessTogether (a crowdsourcing platform with detailed accessibility attributes for mobility, hearing, and visual disabilities). For each source, the authors catalogue the available accessibility attributes, data formats, and completeness rates. The paper then proposes using Semantic Web technologies — specifically Linked Data principles and ontology matching — to integrate these heterogeneous datasets into a unified Linked Open Accessibility Data (LOAD) repository that could be connected to broader resources on the Linked Open Data Cloud such as DBpedia and LinkedGeoData.
Key findings
The survey reveals a stark data quality gap between crowdsourced and government-published accessibility datasets. In Wheelmap, 98.44% of over 421,000 UK location nodes had unknown wheelchair accessibility status, and in Factual, 94.92% of 210,613 UK restaurants lacked wheelchair accessibility information. By contrast, government-published datasets from Transport for London and National Rail had substantially better completion rates — roughly 50% of TfL station attributes were annotated, and National Rail data had less than 2% unknown values for most core accessibility attributes like ramp access (66% yes) and step-free coverage (52% yes). The authors also found that accessibility data is fragmented across systems with no standard schema for accessibility attributes — Wheelmap uses a single wheelchair-accessible field while National Rail tracks 13 distinct accessibility features. The proposed LOAD ontology uses a hybrid integration approach with individual ontologies per data source mapped to a top-level accessibility ontology, using geographic coordinates and shared identifiers (OSM Node ID, NaPTAN ID, Factual ID) to link equivalent entities across datasets.
Relevance
This research highlights a fundamental challenge in physical accessibility: the information needed to navigate the built environment exists but is scattered, inconsistent, and frequently incomplete. For accessibility practitioners, the paper demonstrates that crowdsourcing alone cannot reliably populate accessibility data — government and institutional data publication is essential for baseline coverage. The Linked Data approach proposed here offers a model for how organisations could publish structured accessibility data that machines can integrate and query, which remains relevant as cities and transport authorities continue developing accessible wayfinding tools. The work also underscores the need for standardised accessibility data schemas, a gap that persists today despite growing interest in smart cities and accessible tourism.
Tags: open data · linked data · accessible travel · crowdsourcing · semantic web · wheelchair accessibility · transportation accessibility