We report on investigations into hierarchical phrase-based translation grammars based on rules extracted from posterior distributions over alignments of the parallel text. Rather than restrict rule extraction to a single alignment, such as Viterbi, we instead extract rules based on posterior distributions provided by the HMM word-to-word alignment model. We define translation grammars progressively by adding classes of rules to a basic phrase-based system. We assess these grammars in terms of their expressive power, measured by their ability to align the parallel text from which their rules are extracted, and the quality of the translations they yield. In Chinese-to-English translation, we find that rule extraction from posteriors gives translation improvements. We also find that grammars with rules with only one nonterminal, when extracted from posteriors, can outperform more complex grammars extracted from Viterbi alignments. Finally, we show that the best way to exploit source-totarget and target-to-source alignment models is to build two separate systems and combine their output translation lattices. 1

This paper studies two methods for training hierarchical MT rules independently of word alignments. Bilingual chart parsing and EM algorithm are used to train bitext correspondences. The first method, rule arithmetic, constructs new rules as combinations of existing and reliable rules used in the bilingual chart, significantly improving the translation accuracy on the German-English and Farsi-English translation task. The second method is proposed to construct additional rules directly from the chart using inside and outside probabilities to determine the span of the rule and its non-terminals. The paper also presents evidence that the rule arithmetic can recover from alignment errors, and that it can learn rules that are difficult to learn from bilingual alignments. 1