Weighted Alignment Matrices for Statistical Machine Translation

Current statistical machine translation systems usually extract rules from bilingual corpora annotated with 1-best alignments. They are prone to learn noisy rules due to alignment mistakes. We propose a new structure called weighted alignment matrix to encode all possible alignments for a parallel text compactly. The key idea is to assign a probability to each word pair to indicate how well they are aligned. We design new algorithms for extracting phrase pairs from weighted alignment matrices and estimating their probabilities. Our experiments on multiple language pairs show that using weighted matrices achieves consistent improvements over using n-best lists in significant less extraction time.

• Type: Journal
• Year: 2009
• Where: EMNLP
• Authors: Yang Liu, Tian Xia, Xinyan Xiao, Qun Liu
• Keywords: Alignment Mistakes, Current Statistical Machine, EMNLP 2009, Multiple Language Pairs, Natural Language Processing