STATISTICAL MACHINE TRANSLATION METHOD USING DEPENDENCY FOREST

摘要

PURPOSE: A statistical mechanical translation method is provided to improve a translation performance by applying dependency language models to a translation process. CONSTITUTION: A dependency tree is created by analyzing language pair corpus. A dependency forest is created by combining the dependency tree. Suitable structures are searched for nodes within the dependency forest. When dependency structure is matched with a suitable structure, a translation rule is extracted. The suitable structure is created by controlling the fixing structure of a dependency word.