Home | Site Map | More Information
BackLit is a backtranslator for EBAE that is incorporated into BackNem 2.0.
BackLit uses a special process to backtranslate contracted braille to print. This process, described in the next section, makes possible many new capabilities that have long been requested by teachers and transcribers.
BackLit uses a special process to backtranslate contracted braille to print. BackLit begins its backtranslation process by converting a sequence of braille cells to a sequence of braille letters and braille contractions. This new braille sequence uses a special format that BackLit calls a BuiltWord.
The BuiltWord format is essentially the electronic equivalent of the old format used in braille transcribing manuals to show a braille translation by enclosing contractions in paretheses, e.g. a(ff)(in)(ity). So, for example, if BackLit were backtranslating ASCII Braille A69;Y it would end up with a BuiltWord with the following information:
Of course, a word in the BuiltWord format can be immediately backtranslated to print by simply removing the electronic "parentheses." However, the BuiltWord format can also be used for other purposes. The current version of BackLit has an option that lets the user print out a summary of all the contractions that are used in a braille document.
The next version of BackLit will also allow the user to create a grade-relaxed version of short braille file directly without having to backtranslate it first. (Of course, direct grade-relaxing is only useful for a document that doesn't need to be reformatted.) Full backtranslation and retranslation are required in order to produce a formatted braille file with all the page numbers correct.
BackLit supports the use of a custom indicator, two repeated full cells (dots-123456), to flag literal braille that should be represented in print as simulated braille. This mechanism allows braille fragments to be included in a braille document without causing backtranslation errors.
BackLit includes many features designed to provide unusually accurate backtranslation of EBAE. Moreover, since BackLit will be open source, other developers can make corrections and improvements.
Accuracy of backtranslation is important to sighted adults who are learning braille in order to become teachers and/or transcribers. Accurate backtranslation allows these persons to check their braille transcriptions without having to wonder whether they've made an error or whether they've transcribed correctly but the backtranslator has made an error.
Accurate backtranslation provides an independent check of the correctness of a braille file transcribed from print. (Correct backtranslation is, of course, only a necessary and not a sufficient condition on the correctness of the braille. Backtranslation per se will not show contraction errors that produce correct print. Examples include spelled-out sequences where contractions should have been used and contractions used where they aren't allowed such as in bridging syllables.)
Accurate backtranslation of student work saves TVI's time and benefits braille-using students. Not only are students more independent but if they can count on accurate backtranslation, they don't have to work around backtranslation errors made by indequate software.
Accurate backtranslation can have indirect benefits for braille-using students if their teachers have been in the habit of correcting or ignoring their mistakes on the assumption that any mistakes are the fault of the software rather than the student.
BackLit addresses the problem of backtranslating a dots-36 cell in two ways.
First, BackLit lets the user add words to its exceptions table. Common terms like either/or are, of course, already included in the exceptions table. Unfortunately, there are always going to be new terms that aren't in the table. For example, a recent issue of the ACB Braille Forum referred to an instructor/guide.
Second, BackLit uses a specially-developed algorithm to distinguish the use of the dots-36 cell as an st from its use as a slash. This algorithm considers several trial back-translations based on the various alternatives. It determines the best alternative by comparing the component parts of the backtranslation with the words in a a table of standard English words.
EBAE Rule II.11 for translating words with a portion in italics or capital letters present a challenge for backtranslation. For example, the hyphen is used to terminate the effect of the italics or capitalization indicators where the word is not a hyphenated compound word. This is fine for a human reader but does a computer to distinguish between, say, a fully capitalized compound word and a partially capitalized not compound word.
Literary braille translated BASEball as ,,BASE-BALL where the hyphen terminates the effect of the double caps indicator because baseball isn't a compound word. However, WHITE-COLLAR is translated to braille as ,,WHITE-COLLAR where the hyphen does not terminate the effect of the double caps indicator because white-collar is a compound word. A person familiar with these two words knows the difference. Unfortunately, many compound words are invented words not in a list of standard English words. The best that a computer can do is try removing the hyphen and discover if the unhyphenated word is a standard English word. Since baseball is and whitecollar is not, the computer then knows that the latter must be a compound word.
(In case you are wondering as to how embedded capital letters are used in print, this article analyzing Sylvia Plath's poetry is an interesting example.)
EBAE Rule II.13 for translating words that use embedded hyphens in special ways present a challenge for backtranslation.
BackNem can correctly backtranslate these items. BackNem can, for example, distinguish the use of single-letter whole-word contractions in a compound word such as whip-poor-will [braille :IP-POOR-W] from the use of letters as themselves in stammered words such as w-w-will [braille W-W-WILL] where lettersigns are not used.