Part of the School of Computer Science’s Information Management Group.
The SASWAT screen reading extension has received very positive feedback, and was preferred to a generic base line screen reader by every participant in both an internal evaluation and an evaluation run by an independent third party.
Further details on project outputs are provided below:
The principle objective of SASWAT is to investigate, design, and build a homogeneous mapping framework to support the relating of competing visual streams into a single coherent and mediated accessibility stream such that when automatically applied to a Web document maps from parallel visual to serial audio, thereby enabling visually disabled users to interact with Web 2.0 sites and technologies. By the end of this project we expect to be able to: identify how sighted users interact with Web 2.0 sites and technologies (Work Package 1); identify the impact of Web 2.0 technology and sites on accessibility technology within the context of compound documents and dynamic information streams (Work Package 2); have implemented a machine testable mapping along with any relevant additional non-testable components (Work Package 3); have experimented over this mapping, and in the process of experimentation, built a single, structured accessibility stream for Web 2.0 access technologies (Work Package 4); and finally, have evaluated this new accessibility stream with users, and have verified this evaluation with a randomised controlled independent trial (Work Package 5).
However, before proceeding we must answer a number of novel research questions (directly tied to our five major aims / objectives and work packages):
Successfully addressing these objectives and research questions would advance the state-of-the-art in accessibility and our understanding of Web–2.0 technologies. Its timely nature would mean that our results could be quickly moved to development.
In addition to the core deliverables listed below, all technical reports, data and experimental materials associated with the SASWAT project can be found in the Web Ergonomics Lab data repository.
The SASWAT paper ‘Audio Representation of Auto Suggest Lists’ won the Best Paper Award at W4A 2010.
For a full list of the journal articles, conference papers, book chapters and technical reports authored as part of the project, see the SASWAT Publications page.
SASWAT researcher Dr Andy Brown gave an interview about the success of the project on Insight Radio.
The research has been presented to the public at University of Manchester open days, where initial eye tracking data was collected, and has featured in numerous publications including the Manchester Evening News, PhysOrg and Access IT magazine.
The prototype version of the SASWAT screen reading extension can be downloaded from the software page.
Andy Brown is continuing work on the SASWAT software implementation funded by a grant from Google.
Web Ergonomics Lab student Alex Chen’s PhD is focusing on automatically detecting the widgets that produce dynamically updating content, a process that will greatly enhance the SASWAT screen reading extension’s ability to deal with a wide range of Web pages.
The growth of version two Web technologies is fundamentally changing the way that people interact online. One of the most significant advances behind this new Web is the ability to provide an exciting interactive experience for the user by allowing parts of a Web page to be independently refreshed. A short time ago, navigating the Web was simply a matter of clicking links, moving from one static page to another. Now it is possible to spend a considerable amount of time interacting with a single page through dynamic micro content that update independently (such as an AJAX Widget), without changing the page-address. However for visually impaired users, these pages may be confusing as updates are sometimes not recognised by their screen reader or assistive technology, while in other cases these updates may inappropriately interrupt the user. Visually disabled users, can have real problems accessing Web applications – such as calendars, tickers and suggestion lists – found on travel, entertainment and social networking sites. This is because the technology which converts the visual page to audio doesn’t say when a Web page changes, making much of the Web 2.0 inaccessible to people with visual impairments.
Our objective was to investigate, design, and build a homogeneous mapping framework to support the relating of competing visual streams into a single coherent and mediated accessibility stream such that when automatically applied to a Web document a mapping from parallel visual to serial audio could be achieved. To achieve this objective we undertook fundamental research in the areas of: the cognition and perception of dynamic Web based information; the nature of the new Web interaction / infrastructure model as it evolves; and new Web technologies when applied to visually disabled and sighted users. Addressing each point, we have developed a greater understanding of how sighted users view and interact with dynamic changes to Web pages. We have characterised the range of updates in a taxonomy, and determined how current assistive technologies handle them.
We have completed a series of eye tracking studies documenting how sighted users view and interact with a wide range of dynamic information, both when browsing and completing directed tasks. This has resulted in a detailed understanding of how people respond to dynamic content in a variety of situations, and a model of the likelihood that someone will view dynamic content as a function of its properties. In addition, a survey and online experiment have provided information about the types and prevalence of assistive technologies, and how these different technologies respond to dynamic updates.
The sighted user studies showed that rather than splitting attention between different sources of updating information, people focused on updates they had initiated, and rarely viewed updates that were automatic (such as tickers and animated ads), so the conflict between different sources of updating information is, in reality, minimal. As user-initiated updates generally occur in serial fashion, prioritising these in a serial audio mapping can provide an experience that is similar to that of sighted users, without significantly increasing cognitive load. It remains important to notify visually disabled users about how information is changing, however, so where uninitiated updates occur outside the user’s focus, they are announced by a non-intrusive non-verbal cue.
The user studies showed that attention was allocated according to both the content and the users activity. We have therefore implemented a system as a Firefox browser plug-in, which allows us to monitor the user’s location within a document and his or her activity, as well as tracking when updates occur, how they affect the page, and what their content is. All of this information is used to determine when and how to present each update. Evaluation of the system with 12 blind and visually impaired users found that it is an effective tool, and was preferred by a range of participants to how this type of content is currently handled by screen readers. These findings were confirmed by a second, externally run, evaluation with 10 visually disabled users.
The research work undertaken by the Web Ergonomics Laboratory (formally the Human Centred Web Laboratory), here in the School of Computer Science at Manchester, has transformed our understanding of the interaction requirements of visually disabled users. It was originally thought that sighted users would attend to different visual updates in parallel, however, this is not the case. Sighted users prefer to focus on one user initiated task and complete it before moving to the next. This knowledge enables a better translation form visual to auditory input and removes the threat of the expected auditory cacophony from occurring at anywhere near the levels expected. We have tested our theories buy capturing them as a software tool and presenting auditory content from this tool to blind users. Evaluations have demonstrated that this is an effective tool, and was preferred by a range of participants to how this type of content is currently handled by screen readers.
This knowledge, coupled with an understanding of the particular needs and problems of visually disabled users, have resulted in the creation of a set of principles which suggest how audio interaction with Web pages might take advantage of updates, presenting them in a way which help users achieve their tasks, rather than confuse them. This work has now received a follow on funding grant from Google who wish to see this work brought to a practical resolution in their ‘Chrome’ Web Browser. By using UoM technology in partnership with Google technology we hope to to bring real practical benefits to visually disabled Web surfers while contributing back to the research community which initially sponsored this fundamental work.