MEAT

Multilingual Environment for Advanced Translations

General overview.
Basic classes.
Formalisms.
Components.
Bibliography

MEAT
Multilingual Environment for Advanced Translations
This documentation is available as postscript file from http://crl.nmsu.edu/~ahmed/Meat/meat.ps
Meat, the Multilingual Environment for Advanced Translations, is an application for automatic translation of text in a variety of languages. It has been developed at the Computing Research Laboratory (CRL) at New Mexico State University (NMSU) in Las Cruces, New Mexico. The first application involved Persian-English machine translation for the Shiraz project at CRL. Currently, we are applying Meat for translations from Persian, Korean, Japanese, Spanish, Russian, Serbokroatian, and Turkish. The main design goals for Meat were:

Multilinguality. Meat is supposed to process a wide range of languages. The least obstacle for doing this is the presence of a multitude of scripts for writing text in different languages. Consequently, Unicode is used to represent character data throughout the system.
Declarativity. It is highly advantageous to separate linguistic information from the mechanisms which process it. We do this by using a typed feature structure formalism to represent all kinds of linguistic objects in the system.
Integration. Representing data in an integrated way reduces interface problems between different parts of a complex system. We represent hypotheses as edges in a simplified layered chart.
Modularity. Subtasks of linguistic processing are encapsulated in components which are independent of each other.
Configurability. In order to process several different languages using a variety of tasks, the translation system must be highly configurable. We use application definition files to compose applications from individual components and to supply components with necessary parameters.
The initial version of Meat was implemented by Mike Freider and Jan Amtrup in early 1999. Since then, a number of persons have contributed to the implementation. The people directly or indirectly involved with Meat are: Jan W. Amtrup, Mike Freider, Denis Kamotsky, Karine Megerdoomian, Mohamed Noamany, Remi Zajac, Ahmed Abdelali.
Meat is (c) 1999 by CRL. It is available for research purposes only. For contact, write e-mail to ahmed@crl.nmsu.edu.

Author:: Jan W. Amtrup
Computing Research Laboratory
New Mexico State University
{Updated by Ahmed Abdelali}
{\today}
Version:: 2.1

alphabetic index hierarchy of classes

generated by doc++