Pattern-Based Machine Translation (PBMT) Method is one of machine translation method which performs the syntactic analysis phrase and structure transfer phrase at the same time using bilingual patterns, pattern pairs of a source language pattern and its translation pattern for target language. PBMT used to expand the length of patterns to sentence-length in order to reduce ambiguities in translation. But it brings out the problem that the number of patterns increases rapidly. This paper proposes a model which shortens the length of patterns to phrase-length and reduces the ambiguities in translation by using two-level Translation Pattern Selection Method.
The first step is performed during pattern matching phrase. In the pattern matching phrase, it is most important to find the syntactic head of an argument. In the first step, this problem can be solved by using a hybrid method of examples and semantic constraint with thesaurus. By using the hybrid method, the ambiguities of the syntactic analysis phrase can be reduced and also some proper translation pattern categories of the verb phrase are selected according to the semantic information of the argument. And the second step is performed during pattern transformation phrase. This step tries to find the most natural translation pattern for the verb phrase among the selected translation pattern categories by using statistic information of target language.
By using the proposed model, we could shorten the length of patterns with reducing the ambiguities in traslation. And instead of incorporating the semantic structures of two languages, but by using only the semantic information of each language, the overload to incorporate two semantic structures can be reduced.