Automatically Generating Tailored Accessible User Interfaces for Ubiquitous Services
Julio Abascal, Amaia Aizpurua, Idoia Cearreta, Borja Gamecho, Nestor Garay-Vitoria, Raúl Miñón · 2011 · Proceedings of the 13th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2011) · doi:10.1145/2049536.2049570
Summary
This paper presents EGOKI, a system that automatically generates tailored accessible user interfaces for ubiquitous services based on individual user capabilities. In ubiquitous computing environments, people access services like information kiosks, ticket machines, ATMs, and elevators through their mobile devices. However, the interfaces downloaded to these devices are typically one-size-fits-all, creating barriers for users with disabilities. EGOKI addresses this by combining two inputs: an abstract service description written in User Interface Modelling Language (UIML), and a user capability profile stored in an ontology-based knowledge base. Rather than modelling users by their disabilities, the ontology describes their capabilities across five communication modalities (visual, auditory, cognitive, motor, and speech) at three levels (high, low, null). For each interaction element in the service, EGOKI's Resource Selector chooses the most appropriate media type and resource based on the user's capabilities — for example, selecting alternative text instead of images for a blind user, or high-contrast enlarged images for a user with low vision. The Adaptation Engine then applies structural and presentational adaptations (such as navigation support, heading inclusion, or layout simplification) and generates the final XHTML interface using XSL transformations. The system was developed building on experience from the INREDIS European project on accessible ubiquitous environments.
Key findings
Validation was conducted through two methods: a Barrier Walkthrough evaluation by four accessibility PhD students, and a case study with a blind screen reader user. The Barrier Walkthrough compared EGOKI-generated interfaces against a baseline UCH/Webclient interface for four user categories (blind, low vision, motor impairment, cognitive disability) on an underground ticketing service. EGOKI-generated interfaces had significantly fewer accessibility barriers overall (6 vs 18 for UCH/Webclient). For blind users, EGOKI produced 3 barriers versus 7 for the baseline; for motor impairment users, EGOKI had zero barriers versus 4. Only for low vision users did EGOKI perform slightly worse in one area (losing shape when enlarged). In the case study, a blind user with JAWS 11 completed all four ticketing tasks successfully using the EGOKI interface but could only complete one of four tasks with the UCH/Webclient interface, encountering problems with mouse-dependent events, dynamic content changes not announced to the screen reader, and unlabelled form controls. EGOKI overcame these issues by splitting the interface into multiple static screens with proper heading structure and accessible select controls.
Relevance
EGOKI demonstrates an important approach to accessibility at scale: rather than requiring every service provider to design multiple accessible interfaces, the system generates them automatically from abstract service descriptions. This is particularly relevant for ubiquitous computing environments where users encounter many different services and cannot pre-configure each one. The capability-based user modelling (what users can do, not what disability they have) is a more flexible and respectful approach than disability labelling, and the ontology's 125 capability combinations provide fine-grained personalisation. For accessibility practitioners, the Barrier Walkthrough methodology offers a systematic way to evaluate adaptive interfaces against known barriers for specific user groups. The main limitation is that the quality of generated interfaces depends entirely on service providers supplying rich alternative resources in their UIML descriptions — if a provider only offers text, EGOKI cannot generate audio alternatives. Despite this dependency, the approach represents a practical path toward making ubiquitous services accessible without requiring individual service redesign.
Tags: adaptive interfaces · automatic interface generation · ubiquitous computing · ambient assisted living · user modeling · blindness · low vision · motor impairment · cognitive disability · ontology