Yuri Rubinsky explores use of SGML to generate text for sight-impaired

Distinguished Seminar Series Report

by Mark Crook

Yuri Rubinsky is president and co-founder of SoftQuad, a leader in the creation of software for SGML; SoftQuad's products also include HoTMetaL, an HTML editor. Dr. Rubinsky is the editor of Charles Goldfarb's acclaimed The SGML Handbook and the co-author of the historical novel, Christopher Columbus Answers All Charges. His present areas of interest include improving access to electronic documents and the promotion of standards in the World Wide Web environment. To further that end, he is a member of the Internet Engineering Task Force working group which is developing and maintaining HTML.

The term "access" has begun to have a specialized and somewhat technically oriented meaning in the library community. Specifically, access is used as an alternative to acquisition of library materials. In other words, the library has fulfilled its mission if it is able to get the requested information for a patron even though a particular library might not have it. Access is also used to refer to the ability of a computer to find and retrieve a citation or document in an electronic database. In both cases, access is enhanced by the associated technology. In the first case, interlibrary networking facilitates finding remote sources of information. In the second, a data collection is organized through software and made available through a combination of computer hardware and software.

But what of the human side of access? Is there yet another possible meaning or use of the term as it applies to making buildings and materials accessible to patrons with special needs? Certainly many libraries have led the way in making their facilities handicapped-accessible. What if the patron's disability is visual impairment? Once patrons are inside the building, what do libraries offer them if their needs require alternative forms of typical library materials? Unless the library is specifically oriented toward sight-impaired patrons, chances are that for them, access to the material is virtually impossible.

This problem provides an opportunity to apply technology to facilitate access for the print-disabled. Yuri Rubinsky and his International Committee for Accessible Document Design (ICADD) colleagues have devised a method of electronic document markup and transformation based on the Standard Generalized Markup Language (SGML) that leverages the existing document structure and enables rapid production of alternative forms of a text for the visually impaired. Dr. Rubinsky explained the process as part of the Distinguished Seminar Series at OCLC on Oct. 11, 1994.


In 1989, Dr. Rubinsky received a letter from Jesse Kaysen, formerly of Raised Dot Computing, outlining the way in which she had used SGML as a <169>consistent, easy way" to pour documents into software that could produce Braille output. This achievement was notable because it highlighted the importance of an output- independent language for converting electronic documents into Braille. Specifically, it showed three important points: 1) there was a good mapping from the nonprint aspects of SGML to Braille; 2) the potential also existed to transform the markup to other output manifestations for the visually impaired (e.g., large print or voice); 3) automated transformation of SGML-marked texts could open up a world of documents to the visually impaired. Given that a standard mapping protocol could be developed, an automated transformation could greatly reduce the cost ($1,600 per book produced by manual conversion) and production time (typically eight months for a manual conversion of a print book to Braille). Moreover, the creation of these SGML elements could facilitate the production of large-print and voice-synthesized versions of the text as well.


Dr. Rubinsky said SGML is a language that allows the author to define a markup environment in the flow of a document to tell you something about the document, that is, the intention of the text. SGML is geared for the reusability of context; it is a hardware- and software-independent, U.S. and international standard which has been adopted by industries and supported by vendors.

OCLC senior research scientist Keith Shafer describes SGML as a meta-language for writing Document Type Definitions (DTDs). A DTD describes how a document conforming to it should be marked up. For instance, a DTD will describe which structural tags (elements) may occur in the document and in what order. Simply put, a DTD describes a class of documents. Documents can be validated against a given DTD to see if they conform to the desired format.

A basic SGML text consists of a DTD and the document content; content is composed of markup and data. The DTD defines the valid markup elements for a document type and what they mean. In other words, a DTD is a descriptive envelope for the contents of the document.

The key thing to remember about DTDs is that they allow validation of attributes and attribute values for each element without marking up the document again for a special output application (like Braille). The SGML parser determines whether the document structure is valid, then the output application (print, Braille, Mosaic, etc.) determines what the special markup attributes mean to it.

<!ATTLIST TEXT style CDATA #implied> ]>

This DTD defines a document type article as consisting of two structural parts, author and text. The elements author and text are further defined by the terminal symbol PCDATA which means that the contents of this element is parsed character data, i.e., data characters in the text that are not markup. Finally, the text element has an attribute, style.

In the "document instance," markup tags surround pieces of document content and describe the function of the pieces in the document's overall structure. Attributes modify the elements by carrying additional information about the document section.

<author>Yuri Rubinsky</author>
style=emphasized>This is a SHORT article</text>

The tags in this example are in bold and the attribute is in italic. The resultant printed output given the above DTD and document instance could be: This is a SHORT article Note that the resultant output could appear like the string above. This is because an (output) application could use the style attribute value "emphasized" to actually italicize the text. There is no physical manifestation presumed by the style attribute value.

SGML and the ICADD Methodology

Since December 1991, Dr. Rubinsky has served as a member ICADD, which is developing strategies and techniques for the use of SGML to generate Braille, large-print and voice-synthesized texts. ICADD's work began with three assumptions: 1) the markup technique must be straightforward and simple; 2) only one set of markup--if a second markup is required for nonvisual encoding, it will likely not happen; 3) archival documents must always contain the richest possible markup, thereby further facilitating access to the document. Given those assumptions, the ICADD technical subcommittee's goals were: 1) to make the transform process as automatic as possible; 2) to keep the technique simple; and 3) to reduce the costs involved in making texts available for the print-disabled community. The committee believes it is possible to have creators of DTDs build in the relevant attributes to allow for Braille, large-print, and voice-synthesis from the files encoded for other purposes, as a by-product.

The ICADD 22

The result of ICADD's work is a set of 22 attributes to support basic output formats available in Braille, large print, and computer voice. These attributes describe how elements should be transformed by an application attempting to create mapped output. The ICADD 22 are part of the ISO 12083 (SGML) standard (Appendix A.8: Facilities for Braille, large print, and computer voice), and as a suite they are known as SDA or SGML Disabled Access transforms. In SGML, attributes can appear in element tags like "style=emphasized" in the "text" element tag in the example. However, it is also possible to "fix" the value of an attribute in the DTD so that element tags do not need to contain the attribute explicitly. The ICADD work uses the "fixed" approach. That is, the DTD designer associates fixed SDA attributes and values with each element to be transformed. As a result, anywhere the element is found in the text, it will be uniformly mapped to the ICADD 22 equivalent.

The Production Process and a Success Story

After the markup and validation process is complete, the SGML document is handed over to the document output system, computer software that creates a publisher-determined output form from the SGML document. For example, two companies whose software performs the Braille transformation are Esoterica (Omnimark) and AIS.

Dr. Rubinsky cites the production of his novel, Christopher Columbus Answers All Charges, as an example of the successful use of the ICADD architecture and the SDA transforms. The SDA transforms were used throughout the creation of the work, which meant that as soon as the authors had completed their intellectual activity, the work could be submitted to publishers to create versions of the work for various visually impaired readers. Following the receipt of the electronic version of the SGML text, it took two hours to create the Braille version, 90 minutes to make a voice-synthesized version, and two days to create the large-print version. Dr. Rubinsky is quick to point out that the fact that the work was a novel expedited the transformation process, but he believes that similar gains can be achieved for texts in all subject areas. Clearly, complex texts will take longer to transform, but less than the eight months that it now takes. As a footnote to this publication process, the trade paperback version of Christopher Columbus ... appeared three months after the print-disabled versions.


Widespread application of this transformation process will break down the cost and time barriers imposed by present conversion technologies. It could indeed take a practical step forward in providing universal access to all library patrons regardless of ability or disability.--Mark Crook is consulting systems analyst, OCLC office of research.

OCLC invites applications for Research Fellow program

by Keith Shafer

The OCLC office of research has immediate openings in the 1995<196>1996 OCLC Research Fellow program. The OCLC Research Fellow program brings scientists, educators and administrators with demonstrated research capabilities to OCLC to conduct research with OCLC staff, facilities and data resources.

An OCLC research fellow is expected to conduct research that focuses on problems of significance to the library and information science community; the research need not be specific to OCLC's development and production activities.

Successful candidates must have doctoral degrees or equivalent training in library, information or computer science; applied mathematics; statistics; psychology; or human factors. Experienced researchers will be expected to perform independent research at OCLC, while those at the start of their careers will generally be assigned to one of OCLC's ongoing research projects under the direction of a research scientist.

OCLC expects research fellows to publish research results in the open literature. Publications stemming from research conducted while at OCLC are attributed to the OCLC research fellow; acknowledgment of OCLC's support is required, and coauthorship is expected when OCLC staff make significant contributions to the research effort. The length of appointment for an OCLC research fellow is variable, traditionally six months to one year. OCLC offers a competitive salary and benefits, and relocation assistance is available.

Candidates should submit a letter of interest, including a curriculum vita, date of availability, and specific research interests to:

Dr. Terry Noreault
Director, Research and Special Projects
OCLC Online Computer Library Center 6565 Frantz Road
Dublin, Ohio 43017-3395

The OCLC office of research conducts mission-oriented research to provide the library and information science community with theoretical findings and practical applications.--Keith Shafer is OCLC senior research scientist.

OCLC awards three research grants

by Keith Shafer

The OCLC office of research has awarded three Library and Information Science Research Grants (LISRG) to university researchers for 1994.

"We are pleased to be able to support these projects," said Terry Noreault, director, research and special projects. "University-based research adds an important dimension to our research agenda, and these projects promise findings that should be of broad interest to the library and information science community."

The grant recipients and their projects are:

This research will investigate the effectiveness of different types of hyper-text linkages in two hypertext-based information retrieval systems for bibliographic records. A user study comparing search effectiveness for novice and experienced searchers will be carried out for both systems. Results of the study will have applications in the design of bibliographic-based information retrieval systems by determining how the richness of linkage options and what types of linkages provide the searcher with the most effective search environment.

The purpose of this project is to study end-user understanding of subject headings. The researchers will formulate questionnaires displaying subject headings in different contexts and forms, recruit end users, catalogers and reference librarians in public and academic libraries, and ask them to provide the meaning of subject headings. The findings of the project will give direction for improving controlled vocabularies in the area of end-user understanding. Furthermore, recommended improvements will feature computer-based techniques that could be applied to existing files of subject headings in lieu of expensive, time-consuming, manual editorial changes.

Control mechanisms for intellectual access in a virtual library environment are very important factors. This study proposes to identify basic and important considerations as part of the process of developing such control mechanisms; to explore new approaches in knowledge organization; and to present a model that demonstrates one of the approaches for developing such mechanisms. The research will focus on an examination of the mechanisms to access information sources in the client services area in a prototype Environmental Sciences Virtual Library. It will investigate approaches used by various sources, examine semantically cohesive categories in environmental sciences and related fields, and develop a broad scheme of concept categories. A draft model designed for virtual libraries will be revised and tested through an examination of the collected data. An access control mechanism for the library will be designed and initial work plan proposed.

The OCLC Library and Information Science Research Grant program awards grants of up to $10,000 to help foster quality research by faculty in schools of library and information science. Projects are generally completed within one year, and findings are published in the OCLC Research Report series and in the public domain. Application materials for 1995 are now available. For more information, contact the OCLC office of research.--Keith Shafer is senior research scientist, OCLC.