Towards Optimised Population Sourcing for Web Accessibility Evaluation

Web accessibility evaluation is costly and complex due to limited time, resources and ambiguity. We aim to reduce the number of pages auditors must review, and optimise the evaluation process, by employing statistically representative pages. This minimises a site of thousands of pages to a manageable review of archetypal pages using clustering, and significantly reduces the number of pages an auditor must review as they will inspect a page representative of the others in their cluster. Collectively, these clusters are representative of the pages in the target population. To support this, we introduce a framework of six metrics: coverage, representativeness, complexity, popularity, freshness and accessibility. This framework is supported by an accessibility tool divided into three stages: an initial step focused on crawling and downloading pages to run preliminary clustering. The second step calculates the complexity of the population, and the third stage calculates the sample size and produces a final report. Our approach significantly reduces the number of pages an auditor must review, as they will inspect a page representative of the others in their cluster. Collectively, these clusters are statistically representative of the pages in the target population.

Download paper here