2002 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2002 Table of Contents


Chieko Asakawa and Hironobu Takagi
chie@jp.ibm.com and takagih@jp.ibm.com
IBM Japan Ltd., Tokyo Research Laboratory
1623-14, Shimotsuruma, Yamato-shi, Kanagawa-ken 242-8502, Japan


These days Web content has been becoming increasingly visual and increasingly inaccessible to blind users. Since we believe that the Web is extremely valuable to blind users as a new information resource, we have been trying to make the Web accessible for blind users by developing both server side and client side applications [1][2][3][4].

Recently, we have been developing a Web accessibility transcoding system to dynamically transcode existing Web pages. This system works as an intermediary between a user and a Web server. One of our transcoding features is the automatic insertion of missing ALT texts by using the titles of the destination pages linked to images and maps. We felt that automating this feature would be very useful to reduce the Web authors' workload for improving the Web's accessibility, since missing alt texts are a major problem. We then decided to develop a Web content checking and enhancement tool including this feature for the client side, so that authors could easily start using this method without providing an intermediary server or proxy for transcoding.

Our complete system has five components:
- One component is basically a compliance checker for 16 items necessary for Section 508 compliance.
- A second feature is an enhanced search and replace feature. Site-wide problems found by our site analysis method can be fixed in one step throughout the site.
- A third component handles the automatic insertion of missing ALT texts described above.
- The next component analyzes the content and automatically adds a link to the start of the main content in the page.
- The last component is a kind of Web accessibility wizard that uses an intelligent GUI to help authors with tips for improving Web accessibility.

Currently, when Web authors find problems using Web accessibility checking tools such as Bobby [5], they need to open each HTML file to find and repair the error manually. Our system shows them a browser view with a GUI, so that they do not have to open each HTML document to make the changes. Our goal is to drastically reduce the authoring time for the enhancements to improve the Web's accessibility, while teaching Web authors about the 16 critical Section 508 items during the compliance process. In this paper, we will describe the user interface of our Web accessibility checking and compliance tool.


Figure 1 shows the system organization. An accessibility compliance editor can specify any URL for checking as well as any specific item to be checked from the Section 508 list of 16 compliance items. The system groups each kind of problem so it can ask editors to repair all problems of the same kind or it can simply fix all of the fixable problems automatically.

A screenshot of the GUI is shown in Figure 2. The 508-compliant results are saved to the repair database, which is used later to modify the original HTML documents. Table 1 shows the Section 508 items with our current capabilities. As the first step, our system focuses on insertion of alt texts and the skip navigation links that jump directly to the main content. What follows is a description of these features:

Figure 1. The System Architecture Figure 2. A Screeshot of the GUI

Insertion of Missing ALT Texts

There are two primary ways to automatically provide missing ALT texts.
One is to use the title of the destination page that is linked to the image or maps. Another is to look for an ALT text used with other IMG tags having the same destination or the same image file within the site. We have found that it is common for one image to be used repeatedly for many IMG tags where only some of them have ALT text. Since the image file is the same, we decided to use the ALT text used for the other IMG tags. Figure 3 shows an example of the list of candidates of extracted ALT texts.

The editors can select an appropriate text from the list of candidates, or may want to input a more appropriate ALT text, so the GUI allows for both approaches. The same method is also used to ask editors to input ALT texts for an image when there is no candidate found anywhere within a site. Our system shows them a browser view, so that they can specify which image it is and input the ALT text for the image at once without opening the original documents. Another advantage of the interactive repair is that it can help prevent electing useless ALT texts, such as "spacer gif", "button", "image" and so on.

In such cases, the system warns the editor that it is not suitable as an ALT text and asks the editor to input another text string (or a null string). The list of useless explanations can be modified by the editors. This feature is useful to teach Web authors the importance of useful ALT text and to help support human judgment. When the ALT text of an image is changed, all other ALT attributes in other IMG tags having the same destination or the same image file can be automatically changed throughout the site. This is useful when a frequent link target page changes, so the ALT text describing that page needs to be changed throughout the site.

Figure 3. An example of a candidatefs list

Insertion of a Skip Navigation link

The skip navigation link is very useful, since it allows blind users to skip the unrelated information to the main content at the top of the page and jump directly to the main content. However, these links are still rare on actual Web sites. When there is no skip navigation link in a page, our system can automatically insert such a link at the top of a page. Our system is capable of finding the main content position based on our heuristic rules.

The inserted skip navigation link can be adapted to other pages using the same design template, since our system can detect pages using the same template. This means that our editors do not need to insert the link in every page by themselves. The automatic main content detection, however, sometimes makes a mistake, so we provided a GUI for interactive positioning. When editors notice that it is pointing to the wrong place, they can specify another position in the browser view by using a mouse operation. The system provides both automatic and manual correction methods for the following errors of the existing skip navigation links.

First, it checks if the HREF attribute in a link has a description. If there is no valid URL, it will not work when a blind user clicks on it, and the browser will just go to the top of the page. The system can insert some URLs automatically, or ask editors to select the main content position in a browser view.

Second, the system checks if the described URL exists in the page. If the URL is valid but no such location exists when a blind user clicks on it, an error page will be opened. The system can automatically repair the URL or ask for the correct one.

Third, the system checks if there is a skip navigation link at the top of a page. If the link exists in some other position, it does not make sense, so the system will move the link to the top of the page, either automatically or interactively.


We have evaluated our features for supplying missing ALT texts with an actual Web site having a total of about 15,000 pages. There were about 20,000 linked images and image maps without ALT texts, though there were only about 5,000 destinations. This means that this system can reduce the workload to add the ALT texts to 1/4th of the manual work. For these linked images and image maps, the title of the destination pages were appropriate as ALT texts for most of them, and could be used throughout the site. We are able to say that our methods are useful in drastically reducing the Web authors' workload for improving Web accessibility. Our plans are to implement other features into our tools and flesh out a more complete system. We hope to evaluate all of the features with an actual Web site in the near future and report the quantitative results.


[1] C. Asakawa, User Interface of a Home Page Reader, In Proceedings of The Third International ACM Conference on Assistive Technologies (ASSETS '98), pp. 149-156. 1998.

[2] C. Asakawa, H. Takagi, Transcoding Proxy for Non-visual Web Access, In Proceedings of The Fourth International ACM Conference on Assistive Technologies (ASSETS 2000), 2000.

[3] H. Takagi, C. Asakawa, Transcoding System for Non-Visual Web Access (1) -- Automatic Transcoding, In Proceedings of CSUN's Sixteenth Annual International Conference: Technology and Persons with Disabilities, 2001

[4] C. Asakawa, H. Takagi, One-dimensional User Interface for Retrieving Information from the Web for the Blind, In Proceedings of The 1st International Conference on Universal Access in Human-Computer Interaction (UAHCI 2001), 2001.

[5] Bobby, http://www.cast.org/bobby/

Go to previous article 
Go to next article 
Return to 2002 Table of Contents 
Return to Table of Proceedings

Reprinted with author(s) permission. Author(s) retain copyright.