Linguistically-motivated Tree-based Probabilistic Phrase Alignment
Toshiaki Nakazawa, Sadao Kurohashi · 2008
In this paper, we propose a probabilistic phrase alignment model based on dependency trees. This model is linguistically-motivated, using syntactic information during alignment process. The main advantage of this model is that the linguistic difference between source and target languages is successfully absorbed. It is composed of twomodels: Model1 is using content word translation probability and func-tion word translation probability; Model2 uses dependency relation probability which is de-fined for a pair of positional relations on de-pendency trees. Relation probability acts as tree-based phrase reordering model. Since this model is directed, we combine two alignment results from bi-directional training by sym-metrization heuristics to get definitive align-ment. We conduct experiments on a Japanese-English corpus, and achieve reasonably high quality of alignment compared with word-based alignment model. 1