Machine Translation at Texas

Overview

Jonathan Slocum

Work on machine translation (MT) at the Linguistics Research Center was originally supported by grants from the U.S. government; unlike most MT projects in the U.S., which explored Russian-English translation, the LRC investigated German-English translation. The system came to be called called METAL; whether this stood for "META-Language(s)" or "MEchanical Translation and Analysis of Languages" is knowledge now lost in the mists of time. A personal commentary about this early MT work, by Winfred P. Lehmann, founding Director of the LRC, is found in Machine Translation at Texas: the Early Years.

Funding by the U.S. government for MT work at the LRC ended ca. 1975, though very small amounts "for system documentation" became available a few years later, after all the systems personnel and most of the linguists had left the University. Knowledge of the software and how to run it was all but lost, and in any case major hardware changes at the University were soon to render that knowledge moot, and the system unusable.

Eventually, in late 1978, the German firm Siemens AG stepped in due to a growing corporate need for high-speed translation. (Of course MT would come with human post-editing, because post-editing is standard practice even for human translation at any self-respecting company. Americans who had scoffed at the feasibility of MT due to post-editing requirements were, at the very least, remarkably ignorant.)

Funding from Siemens enabled the LRC to hire new systems personnel in 1979, who delivered a full production prototype by the middle of 1984, to be followed by commercial implementation in C++. A personal commentary about this critical phase of system development by Jonathan Slocum, then MT Project Director at the LRC, is found in Machine Translation at Texas: the Later Years.

As METAL evolved into a commercial product, both for in-house use and for licensing to other firms, Siemens gradually absorbed all development work, and LRC MT personnel either became Siemens employees or contractors, or engaged in MT research funded by other sponsors (such as Hitachi for Japanese/English translation), or found other work. By late 1995, give or take, work on the METAL system at the LRC had come to an end, and general MT R&D did not continue for too very long thereafter.

A lightly-edited personal commentary about the commercial version of METAL by Tommy Loomis, who was the main linguistic software developer from 1987-2001, is found in The METAL System: Commercial Development. As late as 2009, METAL was still being marketed (with credit given to the LRC) on the World Wide Web.