Classics and Religious Studies, Department of
Date of this Version
January 1983
Abstract
If a computer is trying to translate Latin into English, what can it possibly do with the dative and ablative plurals, which are all identical? Or, if the text is prose and lacks the long marks, what is it to do with all those pesky first declension -a ’s?
These are human problems, too, for we have all had the experience of having to slow down while we tried out the different possibilities on a final -a. So what’s good for the computer is also good for the human’s Latin reading speed, and the human’s appreciation of good clear Latin style.
It may be apt to take a moment to relate how the computer gets into this in the first place. With all of Greek literature now machinereadable, and with Latin literature fast approaching that happy state, it seemed eminently sensible to start taking computer science courses, to be able to plug in. Then by early summer, 1981, it seemed clear that SNOBOL 4, better known as SPITBOL (the nomenclature of computer languages is traditionally whimsical), a Bell Telephone Labs product, was a powerful enough language to handle the translation of Latin into English. I decided to go at it.
Machine translation is usually attempted with, for a start, a lexicon. Seeing no charms in the chore of typing a lexicon into a terminal-—that’s a lot of typing—-I decided to have at it structurally, and have the enjoyment, much like that of geometry, of teaching the university’s computer to recognize a Latin clause, recognize Latin forms, and to transform the Latin syntax of suffixes into the English syntax of sequence, in short, to reprint Latin clause by clause with each word (most of the time!) in the position it would have in English. Computer efficiency requires handling the high-frequency material fi rst. Since the first thing that many people asked, on hearing I had started such a project, was “What are you going to do about the datives and ablatives?”, it was surprising how long the problem could be efficiently postponed. When it did become worthwhile to find a noncontemplative guide to the dative-ablative look-alikes, remembering my own recent annoyance at replowing Quintilian’s sentences to sort out the final -a ’s, I added them to the investigation, too.
What follows, then, is what I had to find out to guide the program, and myself, through first declension -a ’s, and the dative-ablative forms. Since I have been feeding the University of Nebraska Computer System Julius Caesar exclusively, the investigation was done in Caesar’s Latin. I presume resultant figures from other authors would differ slightly.
Comments
Published in The Classical Outlook 61 (1983), pp. 8-9. Copyright © 1983 The American Classical League. Used by permission.