Though a fairly touchy subject amongst professional translators, machine translation is a field that has always interested us here at The Lingua File. Machine translation had been theorised before the 1950s but today we'll be looking at one of the first forays into the field.
In the 1950s Soviet-American relations were poor, as they were during much of the late 20th century. The Russian language was of particular interest to the Americans, and though professional translations were available, there were concerns that human translations were subject to political bias and interference.
The concept of machine translation had been suggested as early as the 17th century by philosophers René Descartes and Gottfried Wilhelm Leibniz. However, it was a discussion between Warren Weaver and Andrew Booth in 1947 that suggested that natural languages could be translated via the use of a computer.
Between the late 1940s and early 1950s, several experiments in machine (or mechanical) translation were conducted. However, these experiments were limited, used punched card systems, and were hardly groundbreaking.
Leon Dostert, a translator who had worked with American president Dwight D. Eisenhower during the war and had acted as a liaison officer for Charles de Gaulle, was invited to a conference on mechanical translation at MIT in 1952.
Though Dostert was sceptical of the potential of machine translation, by the end of the conference he was convinced there was a future in the field. He did doubt the capable scope of machine translation and preferred experimental methodologies over theoretical approaches to the field.
Dostert had discussed with several other linguists whether or not machine translation was a viable aim, and following the feedback that it was, set out to complete work in machine translation.
Convinced that a small-scale experiment could prove fruitful, Dostert contacted IBM founder Thomas J. Watson, a close friend, to collaborate. The IBM 701 machine that had been released the year previous was used and the programming was written in machine code, a programming language that gives instructions directly to the machine's Central Processing Unit or CPU. IBM chose Peter Sheridan to complete the task of writing the code for the experiment.
|White-Gravenor Hall, Georgetown University.|
He found help and a collaborator in the form of Paul Garvin, a lecturer from the Institute of Languages and Linguistics at Georgetown University in Washington D.C., which was in fact set up by Dostert himself.
Garvin was an expert in Russian, as well as many other languages. He was born in Karlsbad, Czechoslovakia and had emigrated to the US in 1941. He and Dostert decided to test various expressions and phrases from organic chemistry and a few general phrases for their machine translation.
As decided by Dostert, the lexical database was very small, containing only 250 words and six grammatical rules. However, the aim was to show the application of machine translation when it came to morphological and grammatical problems, rather than provide vast quantities of word-for-word translations.
The experiment was such a success that it was widely published in mainstream newspapers such as the Los Angeles Times, the New York Herald Tribune, and the Washington Herald Tribune as well as scientific journals and publications. The story later found its way into local and regional newspapers and excitement was so high that the authors of the experiment claimed that the problem of machine translation would be solved in a matter of three to five years.
Though the estimate appears to be miles from the truth, the Georgetown-IBM Experiment raised the expectations of machines to translate natural languages and made machine translation a potential solution to the wonderfully beautiful and complex problem of translating languages.