© 1989 by Association for Literary & Linguistic Computing
Articles |
A New Version of the Machine Translation System LMT
IBM Thomas J Watson Research Center NY, USA
Michael C McCord, IBM Thomas J Watson Research Center, P O Box 704, Yorktown Heights, NY 10598, USA
This paper describes a recent revision of the machine translation system LMT in which (a) source analysis is based on Slot Grammar, and (b) there is a large language-independent portion of the system, a kind of ‘X-to-Y translation shell,’ making it easier to handle new language pairs. Slot Grammar makes a systematic use of slots (essentially syntactic relations) obtained from lexical entries for head words of phrases. No phrase structure rules (augmented or plain) are used Instead, there are slot filler rules and separately stated ordering rules for slots. A great deal of the Slot Grammar system is in the shell. This includes most of the treatment of coordination, which uses a method of ‘factoring out’ unfilled slots from elliptical coordinated phrases. The parser (a bottom-up chart parser) employs a parse evaluation scheme used for pruning away unlikely analyses during parsing as well as for ranking final analyses The transfer step is designed so that all of the transfer rules except those arising from lexical transfer entries (which are database-like) are in the shell Syntactic generation uses a system of transformations, written in a formalism involving an extension of Prolog unification The revision includes a new treatment of transformation rule ordering. LMT is implemented entirely in Prolog