Abstract:
People with visual impairments are hindered when accessing information on the World Wide Web (Web) as content is not designed with their needs in mind. Visually impaired users can access the Web through screen readers that use the underlying structure of a page to create a sequential, audio rendering of the content. However, most designers are mainly concerned with how content is presented, rather than its structure and meaning. Consequently, implicit information available through the visual rendering of the content is lost to screen readers and therefore users. To address this problem, tools that transcode Web content into a format more suitable for screen readers have been developed. While these tools have assisted users in accessing Web content, limitations have been identified. Firstly, the approaches taken have either been scalable but inaccurate, or accurate but unscalable. Secondly, the transformations have tended to focus on adapting content to meet the needs of the device rather than the user.
This thesis presents work that addresses both these limitations. SADIe, a content transcoder, was developed that is both accurate and scalable. This is achieved by annotating the Cascading Style Sheet of a Website. The annotations provide accurate transcoding as they identify key elements of the page when applying the transformations. As most Websites typically have one set of style sheets that all pages refer to, the annotations propagate to every page providing scalability. Technical evaluations of SADIe established that it was capable of consistently transcoding a diverse range of Websites. Unlike previous tools, the transformations used were based upon an understanding of behavioural strategies users employ when accessing Web content. A study of eleven users identified forty-eight strategies categorised into six abstract patterns. Transformations based on four of these patterns were incorporated into SADIe. Qualitative and quantitative user studies of the behaviour-driven transcoding demonstrated that the approach can assist users in accessing Web content beyond that of previous solutions.
Full Thesis:
Download Darren Lunn’s Full Thesis
Thesis Advisor:
Simon Harper and Sean Bechhofer
Award Date:
December 1, 2009
Institution:
The University of Manchester
Manchester, UK
Author Contact:
darren.lunn@cs.manchester.ac.uk
Comments are closed.