Accessibility to Scientific Information by the Blind:
Dotsplus and ASTER could make it easy *

William A. Barry and John A. Gardner
Department of Physics, Oregon State University

T. V. Raman
Department of Computer Science, Cornell University

Advanced scientific documents have historically been inaccessible to the blind. The symbolic system normally used to express mathematical and scientific ideas is more complex than the letter after letter, word after word, line after line of regular literature. This leads to difficulty when mathematics is translated into Braille or when mathematics is spoken. Now that more and more scientific documents are available in electronic format, this difficulty can be remedied. We describe two methods for displaying technical documents in a non-visual format: ASTER, a system for reading aloud such documents, and dotsplus, a tactual method of printing such documents.

Both ASTER and dotsplus address the fact that scientific literature is usually presented to sighted readers in a multi-dimensional format. It is multi-dimensional in the sense that the information is contained not just in the words themselves, but also in the positions of the words and in the fonts used in printing the words. This multi-dimensionality allows the reader to browse, recognizing the big structures first and then focusing on particulars. For instance, it is quite easy for a sighted reader to browse through a paper and just read the bold-face section titles; or when reading an equation containing a fraction with a complicated numerator and denominator, the reader can notice at a glance that the equation contains a fraction, then go back and read the numerator and denominator.

ASTER

ASTER, an acronym for Audio System for Technical Reading, is a software application that controls a voice synthesizer and an audio soundboard. It takes electronic documents and reads them aloud using the notion of audio formatting to express multi-dimensional information. This audio formatting is accomplished by several techniques:
  1. Voice synthesizer parameters such as pitch and tone are utilized so that not only the word has meaning, but also the way the word is said has meaning; for instance, the math expression x squared which is printed as x with a superscript 2 can be spoken by ASTER first by saying x then saying 2 in a higher pitch voice, the higher pitch indicating a superscript (in contrast, a lower pitch would indicate a subscript).
  2. The audio soundboard is used to create unique tones and other non-verbal sounds which can be used to indicate the beginnings of sections or paragraphs.
  3. Stereo effects are used to make sounds appear in different parts of space. This is useful for reading aloud tables where columns can sound next to each other in space.
  4. The reader may interactively change the rules governing the way that something is read; for instance, x with a superscript 2 can be read "x squared" or it can be read "x super 2" or it can be read "x 2", with the two being read at a higher pitch. Each rendering rule corresponds to a different view of the object being read.

The most common way that students currently receive technical documents is as an audio tape of another person reading the document. This presents a single view of the document in which the only controls over the reading are the choices of the person initially taping the document and the pause and fast forward buttons on the tape-player. ASTER improves on this technique by allowing multiple views of the document using different rendering rules and by allowing much finer control over the browsing. ASTER's browsing mode allows the reader to step back and forth and selectively read or re-read anything in the document. This browsing along with the notion of rendering rules allows the reader to skim the document, for instance, reading all the section titles or reading only the equations or reading the first few words of each paragraph.

Since it is not easy to grasp and understand long complex objects that are read aloud from beginning to end, ASTER has one particularly useful rendering rule which replaces complex expressions in math equations with a single variable. For instance a fraction with a very complex numerator and denominator can be read aloud as "the fraction n over d where n is such and such and d is so and so". In this way the reader gets an overview of the main structures in the equation and later is given the finer details.

ASTER works by first reading an electronic document and then creating an abstract model of it. This model is display-independent, i.e., it contains no explicit information about how to visually format or audio format the document. It is instead a high level representation of the information which can then be interpreted according to the instructions built into ASTER or according to the instructions of the person reading the document. Each document structure/element - such as a section or a paragraph or an equation or a particular part of an equation, e.g., the numerator of a fraction - is thought of as an object in this abstract document model. After this model has been created, ASTER begins reading aloud the document using a voice synthesizer and an audio sound generator. As it encounters each object in the document model, it reads that object according to the rendering rule for that specific object. These rules, written in the Audio Formatting Language (AFL), control the order in which words are spoken, the voice synthesizer parameters, and the audio soundboard. There are several pre-written rules from which the reader may choose, or the reader may write his own rendering rules to tell the system how to read an object.

ASTER is written in the object-oriented language Common Lisp using the CLOS extensions to Lisp.It is currently implemented on a Sun Sparcstation IPC and a Decstation running Ultrix. It is straightforward to port ASTER to any unix platform for which there is an implementation of Lisp and CLOS and there are plans for this system to be rewritten in C++ so that it can be ported to other platforms.

DOTSPLUS

Dotsplus is a tactile method of printing technical literature for blind readers that incorporates both Braille and graphic symbols in a manner that retains the same structure as a document printed for a sighted person. Some of the more easily recognized symbols such as plus, minus, the division line in fractions, etc. are enlarged and printed as raised images, while Braille is used for alphabetic characters, numbers, punctuation marks and other symbols that are hard to recognize as raised symbols. The placement of symbols and characters is the same in dotsplus documents as it would be for a sighted person - subscripts are dropped below and superscripts are raised above the position of the main character; numerators of fractions are placed above the denominators; in advanced equations, symbol placement such as sum and integral limits, are preserved. Thus dotsplus uses some of the same formatting that is used for visual printing to convey information in a tactile format.

In standard Braille, all characters are in a straight line and that line is denotes where the top and bottom of the Braille cell is located. In dotsplus the characters are not necessarily in a straight line. Thus, it is difficult to distinguish a dropped cell such as a punctuation mark from a subscripted cell of another character. To avoid this ambiguity dotsplus incorporates graphics into the dropped cells by putting a solid line above and below the character. These solid lines tell the reader that he is reading a dropped cell and locate the top and bottom of the cell. There are other changes from standard Braille: for example, the use of the numbers from European computer Braille and the use of 8-cell characters instead of 6-cell characters. (For more details on these changes, see our paper Dotsplus - Better than Braille? in the CSUN 1993 Conference Proceedings.)

Dotsplus documents are produced using a set of Truetype or Postscript fonts and are currently printed using a wax-jet printer that has been modified to print a very thick layer of wax. We pass the paper through the printer twice to get good quality raised images - a very tedious process. However, the printer we use is no longer produced, so we are therefore actively exploring other methods of printing.

Another important aspect of the dotsplus method is the incorporation of graphics into the document. Many technical papers have drawings and graphs in them. In many papers that are Brailled for students, these drawings are left out and the student is instructed to ask the teacher to explain them. In dotsplus these illustrations are incorporated directly into the document. One method we use is to scan them in, expand the drawing roughly by a factor of two, and then use a bitmap image editor to replace any print characters in the illustration with Braille/dotsplus characters. The illustration is then printed out in raised mode along with the rest of the document.

In summary, both dotsplus and ASTER provide useful methods for presenting math, science, and other technical documents. There are, however, a few things that must happen before these two systems can be widely used. For dotsplus, an appropriate printer must be designed; ASTER must be ported to other computer systems. Also, both dotsplus and ASTER depend on the availability of electronic documents in an appropriate format. This leads us to a final discussion of the accessibility of documents.

Documents come in a hierarchy of accessibility. The most accessible documents are those written using a markup language such as SGML or TeX/LaTeX. In these documents, the objects are clearly labeled; for example, a fraction in LaTeX is input as \fraction{n}{d} where n is the numerator and d is the denominator, and the beginning of a section is marked by \section{section name}. In order to print a LaTeX or SGML document it is necessary to convert it into a language which the printer understands such as PCL (the printer command language) or Postscript, a so-called page description language. All printer languages are designed to tell the printer only where on the page to put a character and what font to use when printing it. To a page description language, a fraction is just one character over another separated by a line. Page description languages are inherently visual in nature and contain no information about what they are printing except to someone who can see the page. Only slightly below the printer languages in accessibility is a piece of paper or its electronic equivalent, the bitmap. To make these accessible requires sophisticated optical character recognition software and as yet, there is no optical character recognition software that recognizes mathematical text. Therefore, when creating and distributing documents the most accessible format is a markup language such as Tex/Latex or SGML and the least accessible are page description languages such as Postscript or PCL, or paper. Even an ASCII text is not as accessible as markup language because an ASCII text does not have its parts clearly labeled. Thus, the current versions of dotsplus and ASTER work best with TeX/LaTeX and could easily be made to work with SGML documents. It is hoped that more documents in the future will be available in these formats.

* The Dotsplus research program was supported in part by the US National Science Foundation. T.V. Raman would also like to acknowledge the help and advice of David Gries.