INCREMENTAL ENVIRONMENT FOR SCORED MACHINE TRANSLATION SYSTEMS

In a Machine Translation System a wide variety of data and rule bases are involved. Some of these data bases must be constructed, trained or learned from the well-formed output of the system. Such data bases are sensitive to the changes in the other data bases in the system, and the reconstruction of these data bases will be time-consuming if everything is reconstructed from scratch. In this paper, we propose an incremental strategy· to this problem. The basic operations in constructing an incremental interface are outlined and the associated topics are discussed. An illustrative example will be given to show the basic methodology behind the philosophy of incrementality. Some general consideration and extension about incremental approach will be discussed as well. The same strategy can be extended to handle large data base which is subject to change in a much more efficient way. With an incremental approach, the data bases which are sensitive to changes can be reconstructed by making only local change to the original ones. Thus, the time and cost for reconstruction will be reduced significantly.