Towards real-time and large-scale web accessbility

We rely on large-scale web accessibility evaluations to obtain snapshots of Internet Health and understand trends and behaviours impacting overall web accessibility. Such evaluations are financially and time exhaustive, making the possibility of more real-time measurements of Internet Health infeasible. In this paper, we investigate the impacts of optimising the page selection processes of large-scale web accessibility evaluations. We set out to conduct an automated accessibility evaluation of 1500 websites using the ‘Home+’ sampling method (for each website, we evaluated the home page and all pages linked belonging to the same domain) as our baseline; then compared the agreement rates of web accessibility evaluations on further sub-sampled datasets. Accessibility data was successfully captured on 987 websites. Our findings demonstrate that a strong accessibility evaluation agreement between the baseline and the sub-sample datasets could be reached with a sub-sample of just 20% of the pages, significantly reducing the effort and resources required to conduct large-scale web accessibility evaluations.