音声ブラウザご使用の方向け: SKIP NAVI GOTO NAVI

Web Posted on: Decmber 7, 1998


BrailleSurf: An HTML Browser for visually handicapped people

Djamel HADJADJ
Dominique BURGER
INSERM-Creare U483
UPMC B23, 9 Quai St Bernard, 75005 Paris, France
tel: +33 1 44 27 26 25
fax: +33 1 44 27 34 38
e-mail: djamel.hadjadj@snv.jussieu.fr

Summary

This paper reports the work undertaken in a project aimed at providing services via the Internet to visually handicapped students from primary schools (the BrailleNet project). It examines the functionality needed to browse through HTML when the vision can't be used as the main sensory support. It is assumed that users have no experience in computer applications nor in graphical interfaces. Thus non visual browsing methods should be simple and intuitive as to make its learning and use very easy even for children and kids. We describe briefly BrailleSURF, the resulting HTML browser that has been developed is currently in evaluation in 20 French educational settings involved in the BrailleNet Project. The general lines concerning the architecture and techniques used to implement it are drawn.

Introduction

The language HTML has been created to describe the content and the structure of multimedia electronic documents. Since this language has been popularised by the success of the Internet, it has been becoming a matter of standard used by many software manufacturers and publishers. Thus an increasing amount of documents is now available in an HTML form. In complement, reading tools have been developed so that not only present documents via a computer interface but also facilitate navigation through large document databases over the electronic world thanks to powerful engines embedding search, retrieve and navigate functionality.

For people with visual impairment, this situation renews the old problem encountered in looking for information. The question is not any more to get documents produced from the scratch in a completely different form (Braille, audio, large print) but rather to have an alternative access to the same information sources. According to this new situation, the consortium BrailleNet has been created in France in September 1996 in order to promote the use of Internet in the education of the visually handicapped. This consortium counts user organisations, schools and universities, research labs and industrial companies [1]. Its main achievements during the first year of the project were : the creation of an educational network, the creation of a Web Site containing examples of contents useful for students and the development of a Web browser which is the topic of this paper.

A preliminary survey on the software tools that are currently used to access the Web by visually handicapped users showed that mastering these tools necessitates a long and difficult training which could be a severe obstacle in our project. Let us simply remind that visually handicapped users have to deal with 1) a GUI (the graphical layer of Windows), 2) a browser whose interface widely refers to visual metaphors (Netscape, Internet Explorer) and 3) an additional software translating the content of the screen in Braille or speech (such as Virgo, Jaws, ..). For this reason, the development of a specific browser was undertaken. The results are very encouraging.

The filtering of HTML documents

1. General principles

Internet servers provide information via the unified HyperText Transfer Protocol (HTTP), and use the HTML [6] language to describe the structure and content of the information delivered. Thus, this information is fairly easy to interpret and transform even before to be presented to the user. Existing access solutions for the visually handicapped have taken advantage of this feature [2, 4, 8-10].

Different levels of transformations can be distinguished, for instance simplification (images are removed), clarification and rephrasing (a label can be added before an hypertext link to make it clearly perceived on a Braille bar or when it is spoken by a speech synthesiser [10], restructuring (a summary of links can be provided at the beginning of the document to facilitate rapid access to them [4]).

2. Filtering functions

HTML elements are enclosed in delimiters or tags that clearly indicate the nature of the element. It is therefore easy to trigger a transformation function each time a tag is encountered and to stop it at the end of the element. The process can be applied several times for each transformation. Moreover, HTML is flexible enough to provide a variety of ways to give a semantic content a HTML form. This is the basis of any adaptation method based on processing HTML source code. A transformation database can therefore built that defines the relationships between the various HTML tags and the transformation functions or methods [3]. Table 1 gives some examples of transformations that can be operated on the original sources. These transformation can be operated independently from each other.

Text

  • Braille abbreviation
  • Phonetic value of acronyms for their pronunciation

Headings

  • Prefix for Braille display. Ex. : [H1]
  • Insertion of prosodic markers or voice indicator for speech output

Anchors, Links

  • Bracket for Braille display
  • Prefix for speech output Ex. : "Link" BrailleNet
  • Numbering the links

Images

  • Insertion of an Image indicator. IMG 1
  • Insertion of the caption as a text
  • Insertion of a anchor allowing to by pass the image
  • If the image has a link associated to it, but no comment insertion of the URL address as a link.

Tables

  • Insertion of Table indicator. Ex. : "Table with links"
  • Insertion of the caption as a text
  • Insertion of a anchor allowing to by pass the Table

Frames

  • Creation of a list of links corresponding to the frames
  • Prefix for speech output Ex. : "Frame" Menu

Forms

  • Insertion of an Form indicator. Ex. : "Form with 3 objects"
  • Insertion of the caption as a text
  • Insertion of a anchor allowing to by pass the Form

Table 1 : Examples of transformation functions

The Browsing Interface

In this section we present some examples illustrating concretely these filtering methods applied in building the userinterface of the browser and how different modalities are combined to create a simple and user friendly interface.

1. A multimodal interface

The presentation of documents can use 1) enhanced contrasts, fonts, colours as well as large characters on the screen, 2) synthesised speech, or 3) a Braille display. In fact, all those three modalities can be used complementarily. For instance, the search for an URL address, the downloading of a document produce characteristic sounds. If an URL is not found a message is spoken. When a link is displayed on the Braille display, a short sound is emitted as a complement of the indication given in Braille. Reading a documents can be made in Braille or speech. A single key gives the possibility to switch rapidly from one modality to the other.

2. Different reading strategies

Moreover, different reading strategies can be used, for instance reading the page extensively, reading only the links or only headings and links. It is possible to switch from one strategy to another.

3. The presentation of Links

Links in HTML have an essential functional value. Thus they shall be perceived immediately by the user. When the data are to be read on a Braille display, the links regularly disappear for a very short time. The "blinking" speed can be chosen by the user. This solution can be reinforced by filtering, since the filter can insert brackets juts before and just after the links (see Table 1). To activate a link, the user just clicks on one of the corresponding. If the referenced document is an HTML document the software invokes the HTML interface.

If the reference is an e-mail address, the software invokes an e-mail application. If the document is to be downloaded (images, sounds, software files, ...) an adapted "Save As" Dialogue box opens. When reading with a speech synthesiser, the user has the possibility to stop it. At this moment the focus is on the text chunk being emitted. Then the user presses the Enter Key. This action activates the link Mode, presenting the first link in the chunk. Thus the user can go through the other links using the Up and Down Arrows. Once a link has been selected the Enter Key is used to activate it.

4. Multi-frame pages

The frame structure are transformed into link structures. Each page referenced by a frame, can be consulted separately by activating the corresponding link. At any time, the user can go back to the main page (Frame Page) in order to be able to access to an other referenced page.

5. The choice of a language

Since Internet is basically multilingual, a language key makes it possible to switch from one language to another. Four languages are currently available : English, French, German and Spanish.

Implementation

The browser BrailleSURF has three main modules : A Document downloader module sends requests to Web server using the HTTP protocol. It can also load HTML files from a hard disk, a diskette or a CD-ROM, or, lastly, receive HTML pages generated by software engines on the Web. A Filter module processes the source documents and prepares them before their presentation in Braille, speech or large print display. Adapted documents are then delivered to the Browsing User Interface through which the user can read the document and interact with it. Figure 1 shows the general architecture of BrailleSURF:

The BrailleSURF browser has been developed in the environment of Windows 95 and Windows NT using the engineering concepts proposed in the ActiveX technology [5] (COM and OLE Automation). The programming environment was Visual C++ 4.2 and its extensions (MFC. SDK, APL,...). It has been developed in full compatibility with most of the Braille devices available on the French market with either 20, 40 and 80 cells, and possibly equipped withbuttons making direct pointing and clicking possible on items. Also several speech synthesis system can be used among them a software Speech Engine produced by Elan Informatique.

Conclusion

The approach we proposed for the development of this HTML browser clearly separates the adaptation of the web documents from their presentation [7]. It seems to provide a suitable framework for the development of Internet access products whose main features are full compatibility with current and previous HTML versions, and easy updating according as HTML develops by adding new data to the transformation database

Acknowledgements

This R&D project is supported by a grant of the Fédération des Aveugles et handicapés visuels de France.

Reference

[1] BrailleNet, http://www.ccr.jussieu.fr/braillene t/

[2] Guide to Writing Accessible HTML, http://www.utoronto.ca/ atrct/rd/html/htmlvis.html

[3]Hadjadj D., Agro R. & Burger D., Customising HTML by filtering, 3rd ERCIM Workshop User Interface for All, 1997, pp. 219-224.

[4] Lynx, http://lynx.browser.org

[5] Microsoft, ActiveX Accessibility Conference, Birmingham-UK, 1996

[6]Schwartre J., HTML, Data Becker GMBH & Co KG 1996

[7] Stephanidis C., Access to Graphical User Interfaces by blind people, Concerted Action on Technology and Blindness, May 1991, p.1-65

[8] The Web and Disabled People, http://www.w3.org/pub/WWW/Dis abilities/

[9] Unified Web Site Accessibility Guidelines, http://trace.wisc .edu/docs/html_guidelines/htmlgide.html/

[10] WAB: W3-Access for Blind And Visually Impaired, http://www.inf.ethz.ch/public ations/ea.html

Do you know BrailleNet ? http://www.braillenet.jussieu.fr/education

Dominique BURGER INSERM Creare Université Pierre et Marie Curie B23 9, quai Saint Bernard 75252 paris cedex 05 tel. : + 33 1 44 27 34 35 fax : +33 1 44 07 15 85