The LRC Machine Translation System

Austin is currently developing METAL, a fully-automatic high quality Machine Translation (MT) system. After outlining the history and status of the project, this paper discusses the system's projected application environment and briefly describes our general translation approach. After detailing the salient linguistic and computational techniques on which METAL is based, we consider some of the practical aspects of such an application, including experimental results that imply the system is now ready for production use. Two exhibits are appended: a German original text and its raw METAL translation. (This is not the best translation ever produced by METAL, but it is better than average.) We close by indicating some future directions for the project.