Structured Text, the Key to Information Now and in the Future

Author: George Kerscher ,Research Fellow
Recording For the Blind & Dyslexic (RFB&D)

for presentation at:
IFLA Conference August 28, 1997
Copenhagen, Denmark


If we are to use information successfully, the information must be structured in a way that permits us to find what we want and then to read the information effectively. This paper will briefly describe early use of electronic text to provide information. Secondly, the advantages of using text structuring techniques will be shown. Finally, use of HTML to structure electronic text and audio recordings will be discussed. Overall, this paper will describe possible directions the Internet may take and how libraries for the blind and print disabled can take advantage of these developments.

How the first ASCII E-Text books were structured

The first ASCII books were very simple. They contained no "mark up" or "tags" and used only ad hoc conventions to identify different book elements. Each chapter was a separate file with extra files for the front matter, the table of contents, and the index. Headings and paragraphs were surrounded by two hard returns which are displayed as blank lines. E-Text books in this form were easy to produce and easy to use, but there were some major problems:

Nevertheless ASCII books are widely acclaimed for their utility in the blind community. If you ask a person who is blind how they would like a book, most will say ASCII. Why? Most people understand what ASCII is and it is likely the person will get something they can use. ASCII books are the most basic form of accessible electronic information and most people know how to use them.

Different types of structures

It is clear that ASCII books cannot adequately describe the wide range of materials we use. First the character set is much too limited. The available characters should be able to represent all romance languages and then go beyond to the Asian languages. This requires what are known as "double byte" character sets, because it takes two bytes -- 16 bits -- to identify the various characters. Even the current double byte systems need further extension. In addition, books require structural elements such as headings, page numbers, list items, and tables need to be unambiguously defined.


For textbooks to be more useful, some very simple text elements need to be identified. These text elements give the reader better search and navigation capabilities. These basic elements are:


Dictionaries are structured quite differently from textbooks. The following elements need to be identified:

Many times we find that the components used in producing dictionaries are needed in other books. The glossary of terms, for example resembles a dictionary in its structure.

Cookery Books

While in most cases the elements identified in textbooks and dictionaries can be used to produce a cookery book, there are a few elements which may prove to be useful. For example, a person may which to search for a "main dish" which uses "beef." having the category identified and the ingredients would make the searching easy.

Using SGML to describe the structures

As mentioned above, ASCII does not provide a mechanism for representing the characters used in various languages and it does not provide mechanisms for identifying the elements needed for textbooks, dictionaries, cookery books, and other types of texts needed by a person with a print disability. Fortunately there is already a international standard that meets all these requirements. Standard Generalized Markup Language (SGML) ISO 8879 can accomplish all of the requirements mentioned above.

Several years ago RFB&D started using the simple ICADD application of SGML in our E-Text production process. The International Committee for Accessible Document Design (ICADD) had created an SGML Document Type Definition (DTD) (ICADD22.DTD). Once the data is in a SGML DTD, many tools can be used to transform the information. BookManager is one such tool that has been used at RFB&D. This search and retrieval tool organizes the book into a single file with sophisticated search and navigation options. Currently RFB&D is using a product called "SGML Build" which creates the BookManager book from an SGML input stream. RFB&D makes both an ASCII version and a BookManager version from the same ICADD (SGML) input stream. We can offer our members the ASCII or BookManager version; whatever they are most comfortable with.

I have been talking about SGML for the past seven years and many of you already have heard me talk about the importance of SGML in making documents accessible. Now, the most important version of SGML in the world is evolving as a result of the huge growth of the Internet and the World Wide Web (WWW). Yes, I'll say it again, "Hyper Text Markup Language (HTML) is an application of SGML and is the language of the Web."

The World Wide Web's use of SGML, called HTML

HTML and the web started off using SGML tags indiscriminately. There was little planning -- it just started. It was not long before the web became chaotic. Tags were being invented and few of the developers had a background in SGML nor did they really understand structured documents. They just created a tag and then displayed the information with a certain "look." At this early stage, the developers of the web organized under the "World Wide Web Consortium" known as the W3C. (You can visit their web site at: At this point the W3C made the decision to formalize HTML as an application of SGML. HTML 2.0 was the first version of HTML to conform to the SGML specification. Currently, HTML 3.2 is the formal recommendation; HTML 4.0 is in draft status and will probably be recommended by the end of this year.

The web and the print disabled community are still plagued by HTML starting in such an ad hoc manner. Developers of Mosaic and later Netscape and Microsoft's Internet Explorer confused structure and appearance. This is very important to understanding the complex problems facing persons using the internet who cannot see the screen. The correct way to create HTML and SGML is to separate structure and the appearance of the information. The information should be format and appearance independent. The rapid development of the web and the "get ahead" attitude caused careful design to be thrown out the window. The W3C is trying to correct these problems, but it remains to be seen if the faults will be corrected.

Style Sheets -- Separating Appearance from Structure and Content

The nature of information is format independent. Information can be printed on paper, displayed on a screen, recorded by a human, printed in braille, and many other formats some of which are yet to be invented. The information itself does not depend on the format. Once one accepts this principal, it's easy to understand the direction the web should take. Fortunately the W3C understands this and has developed "Cascading Style Sheets" (CSS to deliver information in a variety of possible formats through the web.

Simply put, Cascading Style Sheets are a place to put formatting information. This formatting information is separate from the structure and content of the web pages. the W3C call these style sheets cascading to indicate three layers of possible style sources of formatting:

  1. Authors defined styles
  2. Browsers defined styles (default)
  3. Users defined styles

Audio Cascading Style Sheets (ACSS)

While visual formatting for on screen display is how most people conceptualize information, information can be presented in auditory forms. the HTML 3.2 specification defines Audio Cascading Style Sheets (ACSS) for presenting audio information. Like standard CSS information the ACSS formats the information using "audio formatting." The first application of this was for persons who are blind, but it is becoming clear that this feature will benefit many others. ACSS is important for:

We can expect that for-profit forces will begin to be active in this arena. These forces will benefit the print disabled community indirectly.

Web Accessibility Initiative (WAI)

The Web as we know it right now presents many barriers to the blind or print disabled user. The list is quite extensive, but there is hope. On April 6, 1997 immediately prior to the Sixth Annual World Wide Web Conference held this year in San Francisco, invited guests were present to witness the launching of the Web Accessibility Initiative (WAI). The WAI is an international activity of the W3C with the goals of making the web usable by all persons the world over. The two major components of the WAI are:

  1. Resolve technical issues associated with the HTML specification.
  2. Educate content creators concerning the correct use of HTML to ensure accessibility.

Nice, France in June saw the first formal meeting of the WAI. The second was held in Boston, USA in August of this year. Representatives from ICADD are participating in the WAI activity. Currently HTML is more powerful and much more common than the early ICADD application and use of HTML is encouraged in favor of the older ICADD application of SGML. The working groups of the WAI are still getting organized and the staff has yet to be hired. A working group has been discussing issues on the W3C list server. Funding for the project is coming from the US government and from the Internet community. Funding is expected to be provided by European sources for activities in Europe. You can learn more by visiting the W3C's home pages.

Some technological developments you should know about

We all know about the use of the Graphical User Interface (GUI) in the world today. You probably know that this development has caused great concern in the blind community. Now more people are using the new operating systems and newer computers. Advances in hardware and software make progress easier. Two developments should be mentioned here that are direct benefits of the newer operating systems. While the GUI presents some barriers, they also present opportunities.

Advancement in Synthetic Speech

We are just about ready to start using "software synthesizers." The older hardware synthesizers will soon become obsolete. Using a Sound Blaster compatible sound card, the new generation of software synthesizers will dominate the market in a few years. Since these synthesizers are software based, we can expect improvements over time in the same way that other software evolves. One of the new techniques for creating these new synthesizers is to create a database of possible human sounds in a particular language. The software then reconstructs the words and sentences from this database of recorded human sounds. The fundamental mechanism for producing the sound is from an original recording of a real human being. We can expect major improvements in synthetic speech from this development. The software will not run on older, slower computers and requires modern 32-bit or higher operating systems. This technological development was not practical until this time in hardware and software evolution.

Using Wireless Communication to Connect to the Internet

We've seen cellular phones develop over the past ten years. This technology depends on a line-of-sight transmitter. This limitation and the low bandwidth cellular technology affords will be overcome in the next generation of wireless technology.

Low Earth Orbit (LEO) satellites are currently being deployed. These communication satellites will allow wireless communication from any mountain peak or valley on any continent on earth. The number of LEO satellites being placed in non-geosynchronous orbit ensures that there will be adequate throughput for everybody's communication needs. This means that it will be possible in a few years to connect to the Internet with a phone no larger than a cellular phone from anywhere on earth. Think about the impact this is going to have on our modern society. What a wonderful opportunity for people with disabilities:

Using HTML for text and structuring audio recordings

HTML, if created properly, can be perfectly accessible to persons with all types of disabilities. The information can be delivered through the Internet, on a disk, or on a CD-ROM. The same browsers can be used for presenting the information from all three types of media. The developing multimedia specifications can be used to deliver human audio recordings, or text, or combinations of the two. So, HTML will support all the types of functions libraries for the blind and print disabled persons require. Now the issue is how to produce the information.

When producing HTML documents specifically for persons with disabilities, it is best to use very simple HTML tags. All the fancy formatting and fancy animations you find on many web sites should not be used. Keep it very straight and simple. Remember the browser does the formatting for you -- all you need to do is markup the document with appropriate tags.

Current thinking on How Books Will be Presented in HTML

The design of novels, textbooks, dictionaries, cookery books, and other reference materials should be considered as work in progress. It is possible to present the information using current technology, but there are components that still need to be developed. For example, a dictionary in HTML can be produced, but it is not easy to go to a certain word using current search software. This type of search software will need to be developed. The design of basic books would look like the following:

  1. The book would be in valid HTML.
  2. The book would be a collection of separate, but related HTML files.
  3. Each chapter would be in a separate file.
  4. Front matter, the index and other logical components of the book would be in separate HTML files.
  5. A detailed table of contents file would be used for navigation. This file would have links to all parts of the book. In other words, each head and sub-heading would have a link from the table of contents to that heading.
  6. There may be a simplified table of contents file in addition to the detailed table of contents. Links would be provided from one version of the table of contents to the other.
  7. There should be links to the previous and next chapter and back to the table of contents in each chapter.
  8. W3C conventions should be used to represent page numbers.
  9. W3C conventions should be used to provide descriptions of graphics.
  10. Use standard HTML structural tags to identify elements in the printed book. For example, Headings <h1>, sub-headings <h2>, <h3>, etc, paragraphs <p>, ordered lists <ol>, unordered lists <ul>, list items <li>, dictionary terms <dt>, dictionary definitions <dd>. The full list can be examined by visiting the W3C web site. The specification can be read there. Please note that most books on HTML are oriented toward highly visual web pages.

As stated, this is work in progress. I hope that some serious guidelines will be produced explicitly for libraries for the blind and print disabled that will address their production needs. It is unclear if the WAI will take responsibility for this, or if the DAISY Consortium will produce such a document. It is clear that good guidelines need to be produced for our use.

Adding digital audio recordings to HTML content

Currently the W3C is working on a multimedia specification. This is intended to allow video, audio, graphics and text to be simultaneously presented on the web. For our purposes, the text and audio components are of interest. Using the specification allows HTML and supporting browsers to play audio recordings over the Internet or from a CD-ROM. The DAISY Consortium is well aware of these developments and is involved in the WAI and the W3C's activities.

In it's simplest form, all that needs to be produced in HTML is the detailed table of contents. This HTML file points to the sound files on the Internet or on the CD-ROM that are played. For example, a book with ten chapters, front matter and an index would have 12 sound files. The table of contents file in HTML would point to the appropriate sound file. The sub-headings in the table of contents would point to the sound file at a time offset in the sound file. In this very simple way, a digital audio book will be created.

Although this is work in progress, it is reasonable to expect that the software tools will be available in the next year to produce books in this way. More sophisticated books can be produced as well. Books could contain full text and full human recordings. The developing software will support this product as well. Books of such sophistication may be the exception in a normal production environment. These books will require more time to produce than a book that has just the text or just the digital audio recording. We may find that books combine the two in a logical division of content. The information that lends itself to human narration would be provided with a human digital audio recording and the information that lends itself to synthetic speech would be provided in HTML text. It will be possible to mix the two media together in one hybrid book.


The Internet and the World Wide Web are the most important information technologies since the invention of the printing press. Enormous amounts of information are available already and much more will be available in the future. If a person does not have access to this wealth of information, they will be at a huge disadvantage in their educational, employment, and recreational endeavors. I would like to see organizations serving persons with disabilities work to ensure the accessibility of the Internet. Information that these organizations produce could be made available directly through the Internet. RFB&D feels that it is our responsibility to assist our members to gain access to the Internet and if the information needed is not accessible directly, then it is our responsibility to do all we can to make that needed information available to them.