Web Accessibility Technology for the Blind

Svetlana Stoyanchev

SLTC Newsletter, April 2009

Web technology uses design and graphics to present information to the user. Visually impaired people use screen readers to access web pages. W3C has developed standards to improve Web accessibility for the blind. However, these standards are not always followed by web developers. Research projects at universities and industry laboratories focus on making the Web accessible with audio browsing.

In the past 20 years, the Internet has made a significant impact on how people access information. Today our use of Web technology includes diverse tasks ranging from travel planning to grocery shopping. Email and social networking sites, such as Facebook, Flickr, and MySpace have become a dominant way of communication. Web presentation relies on graphics for presenting content to the user.

There are 1.3 million registered legally blind in U.S. today. Screen reader applications, such as Jaws are available for visually impaired users to access information on the web. Screen readers present the content of the page sequentially giving audio feedback to help users navigate and follow links. However, screen readers do not reflect structural and visual information of a web page such as colors and fonts. Visual information can be essential for identifying important content on a page. A page may contain menus and links interleaved with the main content. While important content of a page can be easily identified by a sighted user, visually impaired users may face a challenge identifying it with a screen reader.

One approach for making the Web accessible to blind users is the creation of guidelines and standards for audio presentation and encouraging web developers to follow these guidelines. In 1998 the U.S. Government amended section 508 of the Rehabilitation Act of 1973 to address Web Accessibility. The law now requires all Federal Government information technology to be accessible to people with disabilities. The W3C consortium has created rules and guidelines for making websites accessible to the visually impaired. CSS3 speech module defines properties that give control over rendering XML to speech, similarly to how CSS style sheets are used for visual presentation of XML. Accessible Rich Internet Applications, ARIA provides guidelines for annotations to dynamic content and advanced user interfaces. It enhances the HTML's Document Object Model (DOM) with accessibility specific properties. ARIA aims to help screen reader vendors to develop better tools for audio presentations. Firevox is one of the open source tools supporting ARIA W3C standard. Firevox is a voice enabled extension to the Firefox browser. It implements keyboard shortcuts for page navigation and handles dynamic content by presenting users with current information. Firevox supports ARIA annotations for presenting page content. This software was developed by Charles L. Chen who is now at Google working with T.V. Raman on AxsJAX (Access jax) framework for injecting accessibility to AJAX, the technology used for creating live application within web browser. While ARIA enhances DOM with accessibility specific properties, AxsJAX enables the injection of such DOM properties into existing Web applications. AxsJAX defines Content Navigation Rules (CNR). Following Wikipedia's model of user created and enhanced content, CNR can be added to any web page through blogs using Firevox browser. At Google T.V. Raman has also developed a ranking mechanism that gives preference to websites supporting accessibility.

Unfortunately, Web development practices vary widely in their support of the accessibility guidelines. The majority of Web content is still not accessible for visually impaired users:

"A review the front pages of 33 major news websites revealed that nearly 50% automatically refreshed their front pages at certain time intervals, which would make screen readers lose focus if the page refreshed while they were browsing it, 30% contained an automatic slide-show with both images and text, and 18% had dynamic menus and tabs. Furthermore none of the 33 news sites followed the ARIA standard for live regions, which would have enabled them to communicate the type of updates that should be expected to occur." [1]

Researchers at Stony Brook University are working on HearSay, a non-visual browser that allows users to browse web pages using dialog. HearSay is designed to facilitate users' access to important content on a page and provide an ability to navigate the page developed without support for ARIA guidelines. HearSay uses the internal structure of the web page to optimize its presentation to the user. It automatically segments the page into menus, ads, tables, and articles, and converts them into VoiceXML presentation forms. HearSay allows users to navigate between segments using both voice commands and key strokes as input. It uses machine learning techniques to identify the most important parts of the page. User studies have shown a dramatic increase in HearSay users' efficiency in finding relevant information on a web page compared to using a screen reader.

Dynamic Web content - such as form submissions that return the user back to the form with highlighted errors, and template-based web sites that change part of the page as users browse from one page to the next - presents extra challenges.

Similarly, automatic refreshes can cause screen readers to lose focus and start reading the page from the beginning. Researchers at Stony Brook and the University of Washington address the problem introducing a Dynamo-diff algorithm for web pages [2]. Dynamo identifies "new" information on the page and presents it to the user allowing dialog communication. A user has the option of "filtering" old content and browsing between old and new pages. User evaluation on three tasks: 1) a form filling task, 2) responding to a dynamic message, and 3) maintaining context on page refresh have shown that for each of the tasks Dynamo was more efficient than the basic audio browsing algorithm.

Researchers at IBM also work on improving accessibility for visually impaired users. They have developed an accessible internet browser that allows the blind to access multimedia content [3]. IBM's tool allows users to control multimedia playback including starting, stopping, and manipulating the speed of the playback using shortcuts. Many blind screen reader users have developed the ability to process audio information as fast as three times faster than typical speaking rates. Hence, variable speed control is helpful for many blind users.

The current move towards wireless handheld technology also benefits from the accessibility research. This technology can benefit anyone who is using a device with a limited or no screen view. Accessibility researchers at Google today focus on touch-screen phones. They already outfitted a touch-screen phone "with software that speaks much like a screen reader on a PC. Now they are working on ways to allow blind people, or anyone who is not looking at the screen, to enter text, numbers and commands." [4] Other newly conceived ideas include phones that read signs, a technology that everyone can benefit from.

• [1] Y. Borodin, J. Bigham, I.V. Ramakrishnan, and A. Stent. "Towards One World Web with HearSay3". W4A 2008
• [2] Y. Borodin, J. Bigham, R. Raman, and I.V. Ramakrishnan. "What's New? - Making Web Page Updates Accessible". ASSETS 2008
• [3] H. Miyashita, D. Sato, H. Takagi, C. Asakawa "Making Multimedia Content Accessible for Screen Reader Users". W4A 2007
• [4] M. HELFT For the Blind, Technology Does What a Guide Dog Can’t. New York Times, January 3, 2009