Source-side Syntactic Reordering Patterns with Functional Words for Improved Phrase-based SMT

摘要: Inspired by previous source-side syntactic reordering methods for SMT, this paper focuses on using automatically learned patterns with functional words which indicate structural reorderings between the source and target language. This approach takes advantage of phrase alignments parse trees pattern extraction, then filters out those without words. Word lattices transformed generated are fed into PBSMT systems to incorporate potential from inputs. Experiments carried a medium-sized corpus Chinese–English SMT task. The proposed method outperforms baseline system 1.38% relative randomly selected testset 10.45% NIST 2008 in terms BLEU score. Furthermore, just 61.88% filtered obtains comparable performance unfiltered one testset, achieves 1.74% improvements testset.