Robotics Reference
In-Depth Information
ing and acceleration device was suggested that could reduce the search
of a full band to 10 or 15 seconds. In a later model of the machine the
friction between the reading head and the selector was eliminated and
the search speed claimed was further reduced to three seconds.
When he announced his invention, Artsrouni was not thinking of
fully automatic translation and certainly not of high quality translation.
But he did believe that his device could be used for producing
quick, rough translations, as indeed hand-held electronic translators are
used today by millions of travellers. He did not expect his machine
to replace human translators but rather that it could act as an aid to
communication.
Trojanskii's Machine
Petr Trojanskii (1894-1950) also applied for a patent in 1933 for a me-
chanical dictionary for use in multilingual translation, but he went much
further than Artsrouni with his proposals for coding and interpreting
grammatical functions. Trojanskii's plan was to use “universal” symbols
and to develop a complete translating machine. Partly because of the po-
litical tension between Russia and the West at that time, Trojanskii's work
did not become known in the U.S.A. until the late 1950s, but when his
work was “discovered”, the logician Yehoshua Bar-Hillel described Tro-
janskii as the “Charles Babbage of Machine Translation”
Just as Babbage
had constructed an early form of calculating machine and had made sug-
gestions about programming, in both cases using the limited technology
of his day (the mid-nineteenth century), so Trojanskii had described how
a translating machine might be constructed using the electro-mechanical
technology of his own times (the 1930s and 1940s).
Trojanskii's patent described a machine consisting of “a smooth slop-
ing desk, over which moving easily and freely in different directions is
a belt provided with perforations which position the belt in front of an
aperture”. This belt was provided with the words of a large dictionary,
with entries in six languages in parallel columns, much as Artsrouni had
done for four. The operator would locate a word of the source language
(the language
from
which the translation was to be made) and then move
the belt so that it displayed, in the aperture, the corresponding word of
the target language (the language
to
which the translation was required).
The operator would then type in a code indicating the grammatical
category or role of the word in question—codes that Trojanskii referred
.
