ABSTRACT
Up until now, most machine translation (MT) systems have been based on the transfer method, where translation is conducted in several phases such as morphological analysis, syntactic analysis, semantic analysis, transfer and generation. The semantic analysis and transfer phases mainly process words while the generation process combines the translations. Although phrase-transfer rules using syntactic information have also been used, such rules, which are described without strictly defined semantic attributes, sometimes cause unexpected side-effects. When the sentence is divided into the smaller units that are required by the semantic analyser, if there is excessive division the meaning of the input sentences can be changed.
