> > Date: Tuesday, July 01, 1997 6:55 AM
> >
> > Dear Jay Bowks,
> >
> > I received your enquiry concerning Interlingua. You seem to have an
> > artificial language of that name in mind. However, "interlingua" is
> also
> a
> > technical term within the field of machine translation, describing
> one
> > major type of machine translation technique. Machine translation
> systems
> > that operate according to the interlingua principle perform an
> analysis
> of
> > the text (e.g., a sentence) to be translated, building some formal
> > representation of its meaning, and possibly other properties. Then a
>
> > translation into some other language is generated from this
> representation.
> > When no alterations are made to the representation before generation
>
> > starts, the representation is referred to as an "interlingua
> expression"
> > (otherwise the system is a "transfer" system). The interlinguas can
> be of
> > many different forms in different systems, for instance, logical
> formulae,
> > tree structures, or other kinds of graph structures. In other words,
> they
> > are not languages meant for human use, but formal constructions
> suitable
> > for computer programmes. I have developed an experimental system
> which
> > partly uses the interlingua technique, and that must be the reason
> why
> you
> > came across my name. I have no knowledge of the artificial human
> language
> > "Interlingua", however.

Jay Bowks shared your letter to him on our listserv, INTERLNG, devoted to
the use of Interlingua.  That brought to my mind the possibility that the
reason you and other linguists using machine translation call this an
"interlingua translation" is because the man responsible for the
development of interlingua may have described this as a possible use for
Interlingua in a paper presented to the Modern Language Association in
1954, in which he described Interlingua, which is a standardisation of
the international elements in the European languages developed by the
International Auxiliary Language Association (with funding from the
Rockfeller Foundation and other granting sources between 1937 and 1951,
the year Interlingua was released to the public.  Dr. Alexander Gode
was the final director of IALA's auxiliary language project, associated
with that project in several capacities between 1939 and 1951. He was
the one who made the final efforts to bringing the language to reality.
It is a natural language based on standardizing the Latin-Greek-Romance
elements in the European languages according to the prototypic forms
from which these elements have evolved in the modern languages. The
idea was to use English, French, Italian, and Spanish/Portuguese (treated
as if they were for all practical purposes a single language) and German
and Russian as alternates to discover the international words and
forms in the European languages. A word was accepted as in the international
vocabulary if a similar variant with similar meaning was found in at least
three of the base languages. The form on which the variants were standardized
was the prototype, the form that either historically or theoretically the
modern variants had evolved from. Some standardization of derivational
systems was also introduced, because the Latin derivational system was
seen as itself the "grand prototype" on which modern languages based
many of the formations of new words from Neolatin.

Anyway, here is the text from the 1954 paper at the Modern Language
Association in which Dr. Gode discussed Interlingua as a basis for
machine translation:

>Convention of the Modern Language Association of America
>
>Conference on Interlinguistics
>(December 27, 1954, 2:00 P. M.,  Hotel Statler, New York City)
>----------------------------------------------------------------------
>       THE PROBLEM  OF FUNCTION AND STRUCTURE IN INTERLINGUA
>                  Per Dr. Alexander Gode
. . . . (Un major portion del texto es hic delete. Le sequente
         es super le idea de "Interlingua como un intermediari
         lingua de traduction).

>Interlingua claims to be a common denominator among the occidental
>languages.  As a practical tool in the international dissemination of
>scientific and especially medical information it has justified that
>claim to a rather impressive extent. Yet this is only the first --
>though probably the most immediate -- practical application of a
>common-denominator language.
>
>A second phase would be its exploitation for the purposes of
>general-language instruction as also in the teaching of any one of the
>source languages involved in its extraction.  In a teaching situation
>with let us say English as the start and French as the target
>language, Interlingua may function excellently as a third of
>comparison facilitating the recognition of familiar patterns in the
>foreign ones since both would appear as modifications of the
>Interlingua standard.  But a planned exploitation of this potential of
>Interlingua would of course be greatly enhanced by the availability of
>exact data on the comparative function-stricture relationship in
>Interlingua end the ethnic languages involved.
>
>This same prerequisite is considerably more urgent in another
>possible application of Interlingua which so far has received very
>little attention.  Let me hark back for a moment to the very beginning
>of this paper:  If languages were ideal mechanical tools with
>function- structure relations completely coverable by exhaustive keys,
>the theoretical preliminaries to translation by electronic computers
>would be a cut-and-dried affair.  As I see it, the theoretical
>prerequisite of electronic translation is as easily stated as it is
>hard to fulfill. To be able to rig up the translation machine for
>perfect results, all that is needed is a perfect calibration of the
>function-stricture patterns of the departure language against the
>function-structure patterns of the target language   The reason for
>the practical difficulties in this assignment has been variously
>alluded to in this paper. The patterns to be calibrated are not
>stable.  They are as unstable and variable in Interlingua as they are
>in French or English or in any other language.  But if we want to
>connect for instance French and English in an electronic-translation
>set-up, it seems sensible to suggest that the possibility of
>introducing a neutral intermediary of common-denominator
>qualifications should be investigated.
>
>It is quite conceivable that research of the kind proposed in this
>paper will eventually bear fruit in the use of Interlingua as a half-
>way station in MT (as the experts have come to call the new field of
>mechanical translation by means of electronic computers). Instead of
>translating from German to English, one would translate from German to
>Interlingua and from Interlingua to English.  Instead of having to
>calibrate German function-structure relations against those in French,
>those in French against those in English, and so forth in a criss-
>cross maze of connections of all sorts of language pairs, there would
>have to be only back-and-forth calibrations between each individual
>language and Interlingua.
>
>To be sure, even if it can be demonstrated that Interlingua can
>function as efficiently in this assignment as it does in its present
>jobs, it may well be that this can no longer hold true when a language
>pair not comprising a Western language is to be connected by
>mechanical translation.  But rendering Khirgiz texts in Korean or vice
>versa is probably not a very urgent matter.
>
>And anyway  if and when the time comes when the languages of the
>Western world are no longor of prime importance in most international
>affairs, Interlingua will no longer be around either.  As a product
>and physiognomonic expression of Western civilization it seems fitting
>that it should perish with it -- or of course, survive with it and
>flourish.

>>>>>>>>>>>

Stanley Mulaik
Professor of Psychology
Georgia Institute of Technology
Atlanta, GA 30332

[log in to unmask]