Phrase reordering for statistical machine translation based on predicate-argument structure

Mamoru Komachi, Masaaki Nagata, Yuji Matsumoto

Abstract:

In this paper, we describe a novel phrase reordering model based on predicate-argument structure. Our phrase reordering method utilizes a general predicate-argument structure analyzer to reorder source language chunks based on predicate-argument structure. We explicitly model long-distance phrase alignments by reordering arguments and predicates. The reordering approach is applied as a pre-
processing step in training phase of a phrase-based statistical MT system. We report experimental results in the evaluation campaign of IWSLT 2006.