A multiphase machine translation approach, Generate and Repair Machine Tran
slation (GRMT), is proposed. GRMT is designed to generate accurate translat
ions that focus primarily on retaining the linguistic meaning of the source
language sentence. GRMT presently incorporates a limited multilingual tran
slation capability. The central idea behind the GRMT approach is to generat
e a translation candidate (TC) by quick and dirty machine translation (QDMT
), then investigate the accuracy of that TC by translation candidate evalua
tion (TCE), and, if necessary, revise the translation in the repair and ite
rate (RI) phase. To demonstrate the GRMT approach, a translation system tha
t translates from English to Thai has been developed. This paper presents t
he design characteristics and some experimental results of QDMT and also th
e initial design, some experiments, and proposed ideas behind TCE and RI.