Web page clustering for more efficient website accessibility evaluations

Abstract

Despite advances in the legal framework to assure web accessibility people with disabilities still find barriers hindering websites access. The European Internet Inclusion Initiative (EIII) has delivered methods and tools to carry out large scale evaluations of websites. The tools have been used to carry out 180 million tests on 540, 000 web pages to check 1065 websites at a rate of about 7 sites per hour. This paper outlines an approach to reduce the number of web pages needed to compute accessibility scores. The suggested approach relies on machine learning to cluster the web pages according to the barriers detected and to select representative pages for the score calculation. Analysis of the experimental results has confirmed the validity of the accessibility test result as a new feature for clustering web pages, which is planned to be implemented in the EIII website checker tools.