Computerized Braille Production: A Producer's Viewpoint
JOHN BERRYMAN, B.SC., A.A.C.S.
Abstract: Describes the computerized braille production system installed in the Royal N.S.W. Institute for Deaf and Blind Children. Over 500 original volumes of braille have been produced. The system has been upgraded to twice its original capacity and the staff increased to ten persons. The benefits and the disadvantages of the computerized approach are described.
During 1979, the World Council for the Welfare of the Blind convened a conference in London entitled Computerised Braille Production-Today and Tomorrow. The conference was attended by delegates from 18 countries who were either involved in research into computerized braille production or were current users of existing computer technology to produce braille.
Approximately 25 computer installations throughout the world are involved in braille production. Several approaches to the development of computerized braille production systems were described during the conference (Bennett & Hagger, 1979; Hampshire, 1979; Miller & Paris, 1979). These approaches vary in the degree to which the production process is automated. All of them use the text processing capabilities of magnetic medium storage by computer, which facilitates storage, correction of errors, and reproduction of text, whether in print or braille. They differ, however, in the areas of text entry, translation, and braille embossing.
Levels of Automation
Text entry is the process of placing text onto magnetic storage. The least automated, but by far the most common method employed involves typing the text through keyboard terminals attached to the computer. Text entry is controlled by computer programs that vary in the extent to which they facilitate the entry and subsequent correction of text. When the complete computerized system includes translation into braille, the text is entered in alphanumeric format and is keyed through a typewriter-like keyboard; when the system does not include translation, the text is entered in braille through a "Perkins"-like keyboard of six dot keys and a space key.
More automated text-entry techniques make use of the fact that many typesetting systems currently used in the publishing industry use computerized systems and store texts on a magnetic medium. Often this text is in a format suitable for direct entry into a braille production system, bypassing the need to rekey the text. Although some modification of text is unavoidable, this can be automated by using a computer program. The technique is employed to produce braille newspapers and magazines in the United States and in Germany but is possibly most highly automated in the production of braille bank statements at the Warwick Research Unit for the Blind in the United Kingdom (Gill, 1977). The banks provide the research unit with magnetic tapes on which the details to be embossed are stored, and a completely automated system reads data from the tapes, translates it, and embosses the braille statements.
Although optical scanning of the printed page for direct input to computer storage is in its early stages, some success has been achieved with this approach. It will probably become a viable and more widely used method as improvements are incorporated.
Translation computer programs are algorithms employed to convert text from an alphanumeric format to a braille format. Programs specific to different European languages have been developed because each language has its own set of braille contractions. These contractions reduce the bulkiness of braille by representing frequently used words or syllables by a single braille character (cell) or a shorter group of cells. Contracted braille is referred to as Grade II braille (in contrast to the less frequently used, uncontracted Grade 1 braille). Some computer programs are designed to be independent of language and rely on a computer-stored table for each language; the table is referenced by the computer program. Translation programs also vary in the degree to which they provide facilities for formatting the braille.
Automation of the embossing also varies in degree and method. A variety of braille-embossing terminals have been developed and fall into two broad categories: those that emboss paper and those that emboss metallic plates. Most computerized braille production systems have at least one paper embosser, which can be used only for proofreading or for low-volume production. Most emboss on only one side of the paper; some can emboss on both sides by interpointing. Metallic plate-embossing terminals are used when high-volume production runs are standard. (In braille publishing, high volume is 50 to l00 copies for books and 500 to 2,000 copies for periodicals.) The embossed plates are transferred to modified printing presses for press braille production.
Some systems employ off-line embossing techniques. These involve the output of braille-line images onto magnetic cassettes (or punch cards), which are then used to drive modified stereotyping machines to produce embossed metallic plates.
Most computerized braille production systems use a computer dedicated to the task and supplied by Digital Equipment, Data General, or IBM. (Many of these installations are multipurpose: i.e., braille production is only one of the computer's functions.) I know of more than 20 installations: seven in North America, more than 10 in Europe, one in Brazil, and one in Australia.
The Royal Institute's Installation
The Royal New South Wales Institute for Deaf and Blind Children is the oldest children's welfare organization in Australia, founded in colonial Sydney in 1860. The institute provides not only schooling and accommodations for deaf, blind, deaf-blind, and multiply handicapped blind children, but a range of additional services, including medical, dental, and psychological care; vocational training and career education; parent counseling; and library and braille transcription services.
The institute installed a computer in May 1978 for the sole purpose of producing braille.The initial configuration of the Data General Eclipse C330 minicomputer included 96K bytes of main memory, 10 megabytes of magnetic disk storage, one magnetic tape unit, one 60-character per-second printer terminal, two visual display units for text entry, and two line embossing devices (LED-120s, supplied by Triformation Systems, Inc., Stuart, Florida). The application software was purchased from Duxbury Systems, Inc., of Stow, Massachusetts.
Subsequent upgrades in hardware include an additional 32K bytes of main memory, the replacement of the 10-megabyte disk unit with a 50-megabyte unit, and the addition of two visual display units, bringing the total of text-entry keyboards to four (see Figure 1).
Figure 1. Data General Eclipse C330 mini-computer. This figure shows the central processing unit, magnetic tape unit, 10-megabyte magnetic disk drive (later replaced by a 50-megabyte drive), printer terminal, and the two line embossing devices.
Between July 1978 and December 1979, 83 textbooks, novels, plays, and anthologies of poems or short stories (made up of 484 volumes) were transcribed and placed in the magnetic tape library, from which braille copies are produced for sale. Almost half the works were written by Australian authors; about the same number were written specifically for children or adolescents. Almost all the works appear on the New South Wales Education Department's high school syllabus in English, History, or Economics.
Additional transcriptions of about 60 original volumes have been made on request for agencies and associations of the blind. These include magazines, newsletters, agendas and minutes of meetings, information pamphlets, calendars, football fixture lists, operating instructions for tape recorders, and so on.
Production System
The transcription process consists of three parts: (1) entry of text, (2) translation of text into Grade I or Grade II braille, and (3) embossing of the translated text.
Entry of text. Operators type the text at the four visual display units. Text entry is controlled by the Duxbury Text Editor program (Sullivan, 1978), which has numerous advantages over the various Data General text-editor programs available. The editor, a reentrant subroutine that permits editing to take place simultaneously at the four terminals, is "line oriented" and operates directly upon a direct-access disk file. The operators are able to add, insert, and delete lines by referring to line numbers. They also can locate text by searching without knowing the corresponding line number. Alterations can be made on specific lines, groups of lines, or throughout the text. For example, the command "15 to 20 REPLACE TOM BY HARRY" would alter lines 15 to 20 in such a way that "HARRY" would appear wherever the three-character sequence "TOM" used to appear.
The Editor also provides a facility for creating a sequential file from the direct-access file. This sequential file is used by programs other than the Editor. A direct-access file can also be reconstituted from the sequential file. Once a file has become stable, the direct access file is deleted and set up again temporarily for editing only when necessary.
Translation. The second stage, translation, is performed by the Duxbury Braille Translator. This program, written in Fortran IV, is an outgrowth of two earlier programs, DOTSYS II and DOTSYS III, developed in the early 1970s at the MITRE Corporation. The Translator reads the sequential disk file produced by the Editor (or by another program such as a compositor's tape convertor). The input text is in the form of variable-length lines (records) of inkprint text images. The output, another disk file, is the sequence of braille signs equivalent to the input text.
According to its author (Sullivan, 1979), the Duxbury Translator is almost completely table-driven: i.e., details of the translation algorithm are determined by tables read in at execution time rather than by the program. Tables exist for several languages, and there are different tables for American English braille and standard English braille. On the Data General Eclipse, as configured at the institute, the rate of translation is 1,500 words per minute. Translation errors can be corrected either by inserting special "editor's symbols" in the original text or by updating the table, an action that prevents the error's recurrence.
Figure 2. Donald Keeping, supervisor, computer braille service, University of Manitoba, examining braille embossed by the Triformation LED-120.
The Translator expects to receive text just as it appeared in the original inkprint, with a few exceptions: for example, when single quotation marks should have been typed as double quotation marks, even when a quotation is included in another, or when a dash in inkprint should have been typed as two hyphens.
Additionally, the Translator recognizes certain insertions (editor's symbols) in the text that perform specific functions, such as special representation of certain punctuation and print conventions in braille or formatting of the braille pages.
Examples of special representation include capitalization, italics, accent marks, and the correct sequencing of combinations of parentheses, brackets, and italics. Examples of formatting include new paragraphs, new lines, skipping one or more lines, skipping to a new page from other than the last line, tabulation (e.g., for columnar information), hanging indention, and centered headings. The Translator automatically prevents the splitting of words at the end of lines and looks after paging and page numbering.
Other functions that can be controlled through editor's symbols include switching between Grade I and Grade II braille, inserting inkprint page numbers in the braille for cross-referencing, and automatically including a title on every page. If problems arise with the automatic translation of braille, usually deriving from the sound or meaning of the text, symbols that either enforce or prevent the contraction of nominated letter-groups can be included as required.
For the most part, editor's symbols are inserted by the data-entry operators during the data-entry process. When layout of the braille is less straightforward (e.g., braille calendars), the text is subjected to a pre-input step that involves writing the required editor's symbols by hand onto the source text before data entry.
Embossing. The third stage of processing is embossing. The translated text is either transferred directly from disk to a line embossing device (LED-120) or printed out as inkprint braille. At the institute, we transfer the text directly, and then proofread the embossed braille (see Figure 2).
A blind proofreader reads aloud from the braille while a sighted reader checks against the original source. The braille reader also checks for errors in spelling, contractions, and layout. Corrections are indicated on the computer-printed text list, which is returned to the text-entry operator.
Corrections are made by means of the visual display units, and the text is retranslated.The final corrected version is then stored on magnetic tape, from which braille copies can be produced as required.
Each LED-120 can produce 2,000 pages of braille in eight hours. The paper used is continuous, sprocket holed, fanfold stationery weighing 148 grams per square meter.
The computer services staff trims, separates, punches, and binds the brailled sheets into volumes of about 60 pages. The embossing program is flexible with regard to page length and spacing between lines of braille.
Benefits and Disadvantages
Derrick Croisdale (1979), chairman of the World Council's Subcommittee on Computerised Braille Production, attributes the use of computers in braille production to "the inability to recruit, train, and retain manual transcribers." This has certainly been our motivation for doing' so. A system that incorporates a translator enables us to deploy persons with little or no knowledge of braille to produce braille. Only the proofreaders need to know braille. (Because some computer-produced braille is not proofread, no knowledge of braille is required in these instances.) The fact that knowledge of the braille translation/contraction rules resides in the system may be the single most important benefit of a computerized braille production system with translation facility.
A second benefit is an increase in productivity. Although some manual braillists claim to produce as many as 50 braille pages a day, production statistics from the major noncomputerized braille publishers show that the hand braillist's typical output is 20 to 25 pages per day. Often this is the output of a two-person team comprised of a blind braillist and a sighted reader.
At the institute, the average rate of production for each text entry operator has been 40 to 50 braille pages per day -despite the demand for a high degree of accuracy and the fact that many educational texts contain tables, graphs, subscripts, superscripts, and other unusual layouts. (Even in computerized systems that require braille input and do not provide computerized translation, the ease of correcting errors permits a considerable increase in productivity.)
In general, most computers are installed because they perform a task more economically or because the task cannot be performed without them. Whether savings are achieved by the application of computerized methods is a controversial issue among braille producers who use computers. Some are adamant in claiming cost savings. Our experience indicates that computer-produced braille is no more expensive than is braille produced by paid transcribers using the Perkins brailler. The cost of an original page (or master) is about $3.00 (U.S.). Furthermore, it is certainly easier and cheaper to produce additional copies by computer than by the Thermoform copying method.
In the current climate of anxiety about unemployment caused by technological change, it is important to point out that we employ 10 people to produce braille, and have retained our manual braillists to transcribe specialized material such as mathematics.
Perhaps the major disadvantage of the institute's system is that, whereas straightforward textual material can be transcribed easily and accurately, material that requires a more complicated layout can cause difficulties. Although the DuxburyTranslator program includes facilities for almost all eventualities, occasionally the amount of time and skill required by the operators (or proofreaders or editors, depending on how the braille production department is organized) make the process uneconomic. One solution is to produce the difficult passages of text manually. The best solution, however, may be that adopted by the Royal National Institute for the Blind in London (Bennett & Hagger, 1979). Its computer system permits either print-image text input or direct braille input via Perkins-like keyboards. Additionally, specially made visual display units that display braille pages permit posttranslation editing.
Because of Australia's remoteness from suppliers of highly specialized hardware and software, we have had some problems with maintenance. Data General is represented in Australia and provides an excellent computer maintenance service. Triformation is not, and the LED-120 embossers require maintenance at least once a week. This maintenance has been provided by an independent firm in Sydney at an annual cost of about 20 percent of the equipment's purchase price (about twice the normal cost of computer maintenance). Duxbury Systems also is not represented in Australia. Fortunately, however, little software maintenance has been required.
On the Costs of Producing Braille
The most significant item of expense in braille production is the "people" cost, that is, salaries and wages. This applies whether a manual or an automated system is used in the production process. In creating a "master" of a braille page, be it on paper for Thermoforming, on zinc plate for press operations, or on magnetic tape, probably 80 percent or more of the costs have been incurred in paying employees to perform keyboard operations, transcription, and proof-reading. (It must be noted that keying, i.e., typing, and proof-reading are also required for production of printed matter, and there should not be a significant difference between the cost of producing masters in braille or in print format.)
Three approaches have been taken in trying to reduce costs: not paying the people involved in producing braille (i.e., using volunteers); using techniques that increase the productivity of people; and using techniques that eliminate or reduce the need to use people. The first approach is not a real reduction of cost; it simply transfers the burden of cost. It is arguably an unjust solution. Cost reduction by means of increased productivity of people and of greater automation is more fruitful, and computers can play a major role in this. By-passing of the keying of text is important, and it is here that significant saving can be made. As more publishers use computer-based printing systems, and, more generally, as more information is held on computer storage (e.g., teletext, electronic mail, facsimile transmission, automated news, and direct information dissemination), the potential for cheaper braille production increases.
A move away from the specialized braille printing house may also be feasible. It may be more efficient and less expensive to subsidize printed text publishers and to produce braille editions of their publications, than to subsidize the braille publishing houses. If a publisher has a computer-based system, then the addition of a translation program and an embossing device should enable simultaneous publication in print and braille formats.
A second issue is the cost of production of each braille copy of a text. The major cost component of a copy is paper. Typically a copy costs between 2 percent and 5 percent of the cost of the "master" (depending on materials used for binding, etc.).The cost of a braille copy can be 2-5 times the cost of a hard-cover printed book, and 10 or more times as much as a paperback. Compounding this is the fact that economies of scale often do not exist, and the cost of the master must be spread over a limited number of copies.
Over the past five years at least, the cost of paper has risen sharply, and it is likely that this trend will continue. The new technology of "paperless" braille may be the solution. Current difficulties encountered are the cost of the cassette braille recorder/playback machines, the perceived complexity of their operation, and the acceptability of paperless braille to braille readers.
Conclusions
Computerized production of braille is a new, relatively obscure, small-scale activity. According to my estimates, worldwide production is about 500,000 original pages per year. The activity is certainly in a research phase, as indicated by the great diversity of input methods, translation programs, and embossing techniques. At this stage, the fruit of much research effort and expense is little production of braille. However, all those involved in the field of computerized braille production hope that standardization, greater automation, and greater production experience will lead to a quantum increase in production.
References
Bennett, G. F. S., & Hagger, R. J. The R.N.I.B.'s computerised braille system. (Ed.) Computerised braille production-Today and tomorrow, International Conference of the World Council for the Welfare of the Blind,London, England, May 30-June 1, 1979.
Clark, L. L. The future of braille. Braille Research Newsletter, (Warwick Research Unit for the Blind and American Foundation for the Blind), No. 9, April 1979.
Croisdale, D. W. Keynote address. Computerised braille production-Today and tomorrow, (Ed.) International Conference of the World Council for the Welfare of the Blind, London, England, May 30-June 1, 1979.
Gill, J. M. Warwick Research Unit for the Blind, summary of activities 1971-1977. Coventry, England: University of Warwick, 1977.
Hampshire, B. Computer-aided braille transcribing. (Ed.) Computerised braille production -Today and tomorrow, International Conference of the World Council for the Welfare of the Blind, London, England, May 30-June 1, 1979.
Miller, B. E., & Paris, H. B. Compositor's tapes and Library of Congress braille production. (Ed.) Computerised braille production- Today and Tomorrow, International Conference of the World Council for the Welfare of the Blind, London, England, May 30-June 1, 1979.
Sullivan, J. E. The Duxbury braille translation users' manual. Stow, Mass.: Duxbury Systems, 1976.
Sullivan, J. E. The Duxbury editor users'manual. Stow, Mass.: Duxbury Systems, 1978.
Mr. Berryman is manager, Braille Library and Computer Services, Royal New South Wales Institute for Deaf and Blind Children, Sydney, Australia.