Pipelined Multi-Engine Machine Translation: Accomplishment of MATES/CK System

In this paper, we propose a new pipelined multi-engine approach to machine translation, which can take advantage of the previously proposed methods, such as rule-based, example-based, pattern-based and statistics-based methods, and eliminate their disadvantages. Some key new techniques in the multi-engine approach, including attribute knowledge classifications, statistical decision-making, pattern transfer, are discussed. MATES/CK, a Chinese-to-Korean Machine Translation system based on the proposed approach, has been developed.