RE: Machine Translation of K Greek

From: Clayton Bartholomew (
Date: Sun May 04 1997 - 17:47:20 EDT

Larry A. Hartman Wrote:
Just an informational note on machine tranlsation:

        The National Security Agency employs thousands of linguists
across many
dozens of languages. They in cooperation with many research facilities
have been working on machine translating and transcription for several
years. There are many many very technical difficulties which must be
overcome before they employ a package operationally, and to my knowledge
they haven't even begun field testing units yet. If a unit were
employed that didn't have a better than 98 percent accuracy, a strong
possibility would exist in creating international incidents based on
mistranslations. If anyone can program a rigid unserving computer to
accurately translate fluctuating human communications with very minimal
errors (even a language as simple as Greek) I would be very impressed
indeed and would like to know about the product. This post is not meant
for dicouragement to those working this field, but just a means of
informing others of the level of difficulty it is to develope such
capabilities. For the developers, I wish you all the success that the
Lord may grant you.

Larry A. Hartman
Defense Language Institute Alumnus
Department of Arabic Studies

Larry A. Hartman, is correct, total machine translation hasn't arrive
and isn't going to soon.

The US Gov. has been dabbling rather heavily in machine translation for
over thirty years. The first research I saw on automated Russian -
English translation dated from the early or mid 60's. This field is a
major playground for the artificial intelligence community and has
produced trainloads full of unreadable publications which no human could
translate much less a machine.

On translation NT Greek by machine. This is how I became interested in
Greek. I was working on building syntax scanners while reading Chomsky,
F. de Saussure and all those people in the early 80's when I became
rather tired of English and pulled E.V.N. Goetchius off the shelf and
started building scanners for NT Greek clauses. Greek is a more
interesting problem because of the flexibility of word order. At that
time I had the advantage of working daily with artificial intelligence
consultants who could answer questions about special problems.

As long as you stick to morphology and syntax, experimenting with
machine translation can be rewarding. Once you get involved in the next
layer up in the information model, semantics, machine translation is
going to cause you a lot of grief.

This is a specialized subject, and I'm not a specialist. If anyone wants
to find out what is possible or impossible, translating NT Greek by
machine, the people to talk to are at the International Linguistics
Center, in Dallas Texas. These people have spent a lot of time working
on this problem. I am sure that some of their people are on the B-Greek
list if we could only get them to come out of hiding.

Clayton Stirling Bartholomew
Three Tree Point

This archive was generated by hypermail 2.1.4 : Sat Apr 20 2002 - 15:38:14 EDT