Transcription of Braille Textbooks by Computer

BRIAN J. COX, M.Ed., DOS, FAAO

Mr. Cox is coordinator, Resource Centre for the Visually Impaired, Schools Department, Ministry of Education, Vancouver, Province of British Columbia.

Resource Centre for the Visually Impaired, 4196 W. Fourth Ave. Vancouver, British Columbia, Canada V6R 4J5.

Abstract: This article discusses the crisis generated by increased demands for braille textbooks that match print versions in quality of format and graphics. One way of solving the problem is to improve the quality of transcriptions produced by computer. The author reviews the current limitations of computerized braille transcription and points out the existing potential for improvements.

Computerized translation of print texts into braille has been possible for 10 to 15 years and is used in some braille transcription centers in Canada, the United States, and elsewhere (Haynes & Siems, 1979; Keeping, Doyle, & Fotier, 1979). The overall accuracy and sophistication of computer transcriptions, however, has never met all the standards set by the Braille Authority of North America (BANA) (American Association of Workers for the Blind, 1977). These shortcomings are especially noticeable in textbooks that contain complex language and require proper treatment of format and graphics.

These books must be transcribed into grade 2 braille, using all standard contractions (approximately 190) and the various exceptions to the rules on contraction based on common usage. These rules and exceptions are different for different languages and, sometimes, different for the same language in different countries. The standard used in British Columbia's schools is English Braille, American Edition (AAWB, 1980), as adopted by BANA. This code is revised continually: for example, the 1980 revision not only reflects changes in usage but accommodates some problems involved in computerized transcription.

The formatting of braille, like the formatting of print, facilitates the reading and understanding of the content, but it cannot replicate print formats for three reasons. First, braille requires at least twice as much space to present the same content. Second, efficient tactile scanning and sequencing requires boundaries, directions, and cues that are unnecessary in print. Moreover, graphics often must be placed differently in braille versions. Third, the code and format rules for braille reflect a preoccupation with saving space that is not paralleled by current computer software. In fact, in the computer transcription of braille, content is sometimes scattered in a way that obscures the meaning. Although BANA's standards, as set forth in Code Book of Braille Textbook Formats and Techniques (AAWB, 1977), should not be viewed as completely inflexible or beyond orderly or sensible revision, careful attention will be required to provide optimum readability through good formatting without creating unnecessarily difficult challenges to computer transcription.

Pictures, illustrations, graphs, maps, and the like do not convey the same meaning (and sometimes convey no meaning at all) when transcribed from print to a tactile (raised-line) format. Thus, the transcriber or creator of raised line drawings must have considerable imagination, experience, and skill to produce an effective tactile substitute for a visual original. The symbols used in mathematics, music, and computer notation present special problems when developing software for computerized transcription.

Although editing, proofreading, and correcting of print input does not assure that errors will be eliminated in braille, pre-editing of the print by a braillist who has the ultimate braille in mind can substantially reduce errors in braille. But even then, some proofreading of the translated braille is essential before final embossing to eliminate ambiguous language, to correct coding for broad contextual features, to handle line and page runovers, and to position graphs, headihgs, and so forth correctly.

This article addresses the general issue of computerized versus manual transcription of braille as well as the specific issue of whether textbooks that meet code and format standards can be produced by computer. The information is based mainly on a study conunissioned by the Project Planning Centre, Ministry of Education, Province of British Columbia, and conducted by the British Columbia Systems Corporation (Davies & Schmidt, 1982).

Historical Background

Initially, texbooks were transcribed into braille by volunteers serving special schools and government or private agencies for blind people. In most cases, the transcription was tailored to the needs of individual students. In other words, the brailling was done on an ad hoc basis in the sequence and at the pace in which the student used the book. In many cases, only sections of a book that were needed immediately were transcribed, and the complete text was never produced. Finally, because the transcribed material had limited use beyond that of the original student, the student often kept the original master transcription.

As these personalized transcription services grew and spread, the volunteer transcribers formed groups that cooperated with support agencies for the blind and attempted to make their work accessible to more than one student. A good example was the American Printing House for the Blind, established in this way over a century ago. This kind of cooperative, systematic effort did much to coordinate the transcription of braille books and to catalog and duplicate them.

Despite all these moves toward coordinated effort, manual transcription of braille is still a slow and labor-intensive task. Thus, students never have all the braille books they need. In recent years, this low production capacity has been aggravated by the following factors:

  1. Production costs have increased radically as the number of volunteer braillists has declined and paid, full-time braillists have taken their places.
  2. Countries such as Canada increasingly specify national editions of textbooks rather than similar texts brailled in countries such as the United States.
  3. The sophisticated graphics and formats used in modern textbooks are extremely difficult to replicate in the tactile mode.
  4. The integration of blind students into regular classes has increased the need for braille textbooks that match the broad range of textbooks used by sighted students. This is a problem of major proportions in British Columbia, where the total school enrollment is less than 500,000 and only 60 students use braille.

The trend toward integration has had a significant impact on the need for more and better braille for two reasons: First, blind students must follow the curriculum their sighted classmates follow and use parallel learning resources (even edition dates become critical). Special schools for blind students should be capable of adapting existing braille materials as well as creating new ones. However, this capacity is severely restricted in integrated settings. Second, blind students in regular classes do not always have a braille-reading teacher or aide at their elbow. Therefore, errors or ambiguities in the braille text may not be detected and dealt with inunediately.

Manual Transcription of Braflle

Most manual transcription of textbooks is done on mechanical braille-writing machines that produce an embossed paper "master" on one side only. If a large number of duplicate copies are required, these masters can be made on zinc plates for double-side embossing.

The cost of manual transcription varies from $ 2.00 to $ 10.00 per page, depending on whether the braillists are paid staff or volunteers, whether the content is simple or complex, and what standards are set for transcription and proofreading: BANA's standards for coding and formatting are usually followed, (AAWB, 1980). Other variables that affect the cost of transcription are the number of print pages and the number of braille pages required per print page. (The usual rate is two braille pages to one print page).

Because the number of copies required is usually extremely small, duplicate copies are usually made by "Thermoforming"-making a copy of each paper master manually on a plastic sheet. Centralized transcription agencies that produce large numbers of copies use zinc-plate masters and automated mechanical embossing, and duplicate copies usually cost about 25 cents per page.

Canada has fewer than 100 braillists who are capable of transcribing textbooks properly. Of these, fewer than 20 percent are full-time, paid personnel. A skilled braillist who adheres to BANA's standards produces, at best, 100 pages in a 35-hour week. Usually, however, a full-time braillist needs at least four months to transcribe a 500-page textbook. But most students have far less than four months of lead time between the date when they discover they need a textbook and the date when they must begin to use it. Because requests for transcription tend to be clustered around the time when courses begin, manual transcribers are soon overloaded and must scramble to keep up with students' demands. This problem is often compounded by an instructor's legitimate decision not to follow the sequence in which topics are presented in the text.

In today's economy, any skilled labor intensive task is expensive. Moreover, the advantages of freeing skilled braillists from the slow manual procedures of brailling to other jobs such as editing are obvious. Even when volunteers are used to a significant degree, the cost in person hours is considerable and should be avoided, if only to divert volunteer efforts to special materials prepared by teachers and to other personalized work.

Finally, braille is bulky. A print book that measures 10 x 6 x 1 inches (25 x 15 x 3 cm) could be 10 separate volumes in braille and be 40 times greater in bulk (e.g., a set of print books that would fit on a three-foot shelf would, in braille, require an entire wall of shelves).

Alternatives to Braille

In a situation where the demand for braille texts exceeds the supply as well as the technical and financial capacity to produce them, alternatives to braille (such as audio books) should be considered. Despite the general philosophical argument that braille is the only real way to make a blind person genuinely literate, there are also the practical arguments that audio tape indexes are inferior and that readers cannot select pages rapidly and thus review or scan at will.

The Optacon is sometimes regarded as a replacement for braille because it instantly converts print to a tactile image. However, even the most well-trained Optacon users can read at less than one-fifth the speed at which braille readers can read, which is too slow for the requirements of normal study.

The Kurzweil reading machine instantly converts print to synthesized speech. Although the reading (listening) speed is acceptable, the device creates problems of varied access (not page by page) to content, as is the case with audiotapes. Moreover, for now, at least, the price and size of the device make it an impractical alternative for most braille users.

Improvements in optical and electronic magnification (e.g., closed circuit television magnifiers) have reduced the number of people who must use braille by allowing many visually impaired people to become effective print readers. However, a significant number of visually impaired people still lack an adequate alternative to braille. Therefore, the solution may be to improve the computerized transcription of braille.

Computerized Transcription of Braille

Inputting, processing, and storing print data for braille transcription generally follows certain standard procedures. Print is usually entered into the computer for transcription into braille using a key-to-tape or key-to-disc device and a visual display, on which the text can be verified and justified before it is translated into braille.

A skilled braillist can input braille directly on a keyboard containing keys that represent either the six braille dots or the single-cell pattern of grade 1 braille. Because the computer is not required to transcribe the material, the braillist is in a position to use the computer's editing and justification capacity and to store the material in machine-readable form.

The central processing units (CPUS) of most systems are standard or "mini" computers, which have a relatively large processing capacity. The translation system are in high-level language, such as FORTRAN or COBOL (Bennett, 1979). The process uses a table of rules to encode into braille, usually with enough context sensitivity to meet or approach the coding requirements for grade 2 braille (Davidson, 1979). The capital costs for most of these systems are substantial-hundreds of thousands of dollars. Although print can be edited, proofread, and corrected on the visual displays of most systems, the same is not true for braille.

In the few cases where this capacity has been incorporated, the CPU must have a substantial amount of extra storage capacity (25K bytes) to support it. Thus operators must either anticipate the special editing requirements before they key in the material or proofread a printout and request a complete rerun, including a repeat key-in of the material. (Sometimes, several reruns are required).

Furthermore, because the printer operates at a much slower rate than the CPU, the two pieces of equipment must be buffered from one another. This is done by storing the material on tape or a disk. The first braille printout (for proofreading) is in print or embossed paper form. The final printout is on embossed paper if few copies are required and on zinc plates if large scale duplication is needed.

Generally speaking, braille can be produced twice as fast by computer as by manual methods. Computer transcription also offers the potential for even more production capacity because the pool of key-in typists is not limited in the way that the pool of braillists is.

None of the existing systems meet BANA's standards for the coding and formatting of textbooks. Nor do they meet the requirement that braille texts must convey content as effectively as their inkprint counterparts do. Rather than reflecting the programmers' failure, this reflects a conscious decision not to invest the additional time and money required to meet these standards. With sufficient editing and rerun cycles, the standards could be closely approximated, but probably at the cost of the time and money required for manual transcription (Mundy & Droege, 1979). However, current operating systems can produce satisfactory braille of straightforward literary materials such as novels and magazines and newspaper articles (American Association of Workers for the Blind, 1980).

Despite their shortcomings in handling sophisticated material, current operating systems enable skilled braillists to produce more braille than they could manually and to devote their time to difficult transcription work or to personalized transcriptions for individual blind people -work that would otherwise be delayed or left undone. Furthermore, because one braille volume can be stored on one standard cassette tape, or 60 volumes can be stored on one standard tape reel, these systems reduce the amount of space required to store transcription masters and thus eliminate a longstanding problem in the archiving of braille. Finally, they produce masters that are machine-readable and thus are more amenable to editing, revision, and updating, and they improve the person-per-hour cost of production, which could result in substantial savings.

Potential for Improvement

Since the entire field of computer technology is advancing rapidly, new possibilities for better braille transcription are appearing all the time. But because capital and other start-up costs are always involved, the new technology will not be implemented unless the need is clear. In fact, considerable improvements in existing systems have been possible for years; they simply await sufficient demand. Apparently, the need for better computer transcription is developing unevenly; it is particularly intense in situations where blind students have been integrated into regular classes.

Improvements in software offer the biggest gains in money or time savcd. For example, translation can be improved by revising existing translation programs or by writing new ones for the latest equipment. Such programs could reduce the need not only for corrections and reruns but also for pre-editing by skilled braillists. improved software could also solve some of the translation problems inherent in the special codes required for mathematical, musical and computer symbols.

Editing of braille translations on a fully controllable visual display (despite heavy demands on random-access memory) will become less and less of a problem as more powerful low-priced computers become available. Full-page display of braille translations on a visual display with fulll capacity for addition, deletion, justification, and so forth would simplify the proofreading and correction procedures and reduce both braillist and computer time in the process. In fact, even without computerized translation, the ability to key-in braille rather than print would give the braillist all the advantages of a word processor.

New hardware also offers some significant possibilities for improvement. Most notable is the new generation of more powerful microcomputers, which hold promise, not only of reducing capital costs and space requirements but also of increasing the likelihood of more decentralized transcription facilities. Some software has already been developed for braille transcription on microcomputers.

Accommodating the computer to the code is not the only possible direction of change. In situations where it is possible to achieve greater consistency in the code and format rules without loss of clarity or conciseness of expression, compromises that facilitate translation by computer are worthy of consideration. In this regard, it is interesting to note that some recent official revisions of the rules reflect an accommodation to computerized transcription.

Although the need for improvements in printouts is probably less critical, one significant improvement may derive, in a secondary way, from the widespread use of computerized transcription. "Paperless" braille-reading systems depend on direct conversion of machine-readable braille to a braille form on an electronically operated tactile board (Rose & Rose, 1979). Paperless braille, once machine readable masters are available, offers the following advantages: (1) compactness, not only in central archiving but in the user's storage requirements - e.g., one standard cassette tape would equal one embossed-paper volume, (2) the combination of audio and tactile information on the same cassette to communicate information difficult to convey in the tactile mode, (3) greater reliability and speed of duplicating electronic tape, (4) ease of creating and editing braille, and (5) inexpensive, long-range electronic transmission of braille.

Finally, current alternatives to typing into a computer are not practical: optical character reading (OCR) is not yet errorfree, and publishers' composition tapes (when available) are cluttered with print formatting commands.

Summary

The demand for braille textbooks currently exceeds the capacity of manual braillists, and the integration of blind students into regular classrooms will increase the demand for both the quality and quantity of textbooks even more. As yet, however, there are no adequate alternatives to braille.

Computers can produce more braille in less time, reduce the storage problem, and provide "paperless braille." But current systems do not meet the standards as set for manual transcription of textbooks that require specific formats or special codes. For braille to be effective, high standards of encoding and formatting must be met, and the formatting and editing of print does not assure that braille will be formatred and coded adequately when translated by computer. The technology is available, however, to develop hardware and software that meet these requirements in an economically feasible manner.

Some microcomputers are already powerful enough to handle software that would produce acceptable braille textbooks from print keyed in by a typist. Furthermore, it is feasible to use computers as braille word processors - i.e., braillists would be able to edit and format material keyed in as braille.

Finally, simplifying BANA's coding and formatting rules would facilitate computerized translation of print into braille without loss of clarity and conciseness.

References

American Association of Workers for the Blind (1977). Code book of braille textbook formats and techniques. Louisville, Ky.: American Printing House for the Blind.

American Association of Workers for the Blind (1980). English braille, American edition (Rev. Ed.). Louisville, Ky.: American Printing House for the Blind.

Bennett, G. F. S. (1979). The RNIBs computerized braille system. Paper presented at the international Conference on Computerized Braille Production -Today and Tomorrow, World Council for the Welfare of the Blind, Committee on Cultural Affairs, London, England.

Davidson, I. (1979). Braille transcription by computer. Paper presented at the Intemational Conference on Computerized Braille Production-Today and Tomorrow, World Council for the Welfare of the Blind, Committee on Cultural Affairs, London, England.

Davies, D., & Schmidt, J. (1982). Ministry of Education computer braille transcription project.- Preliminary report. Report submitted to the Project Planning Committee, Ministry of Education, Province of British Columbia, Canada.

Haynes, R. L., & Siems, J. R. (1979). Computer translation of grade 2 braille. Unpublished paper, American Printing House for the Blind, Louisville, Ky.

Keeping, D., Doyle, M. S., & Fotier, P. A. (1979). Research for the blind at the University of Manitoba. Paper presented at the International Conference on Computerized Braille -Today and Tomorrow, World Council for Welfare of the Blind, Conunittee on Cultural Affairs, London, England.

Mundy, G. W., & Droege, M. F (1979). The Clovernook system. Unpublished paper, Clovcrnook House School for the Blind, Cincinnati, Ohio.

Rose, S. B., & Rose, J. B. (1979). The Rose display reader. Unpublished paper, Rose Associates, Falmouth, Mass.