Developing a Chinese Module in UNITRAN

This paper will share with the readers our experiences gained in a project of translating Chinese to other languages with Principle Based Machine Translation (PBMT). UNITRAN is a prototype system developed in MIT which translates simple sentences among English, Spanish and German. Based on Government Binding (GB) theory and Lexical Conceptual Structure (LCS) theory, UNITRAN serves a good model for applying GB and LCS to achieve the principle based machine translation. We have tried to put Chinese into the system. Now the system can translate among the four languages properly. In the following sections, we will first introduce the basic idea of PBMT. Then we briefly explain how UNITRAN translates Chinese to English. Our major focus will be the Chinese language parameter setting. Some GB parameters will be discussed in certain detail. And finally, the last section will discuss the merit of the PBMT and problems arise from this approach.