Using reordering in statistical machine translation based on alignment block classification. Internal Report

Statistical Machine Translation (SMT) is based on alignment models which learn from bilingual corpora the word corre-spondences between source and target lan-guage. These models are assumed to be capable of learning reorderings. However, the difference in word order between two languages is one of the most important sources of errors in SMT. In this paper, we show that SMT can take advantatge of in-ductive learning in order to solve reorder-ing problems. Given a word alignment, we identify those pairs of consecutive source blocks (se-quences of words) whose translation is swapped, i.e. those blocks which, if swapped, generate a correct monotone translation. Afterwards, we classify these pairs into groups, following recursively a co-occurrence block criterion, in order to infer reorderings. Inside the same group, we allow new internal combination in or-der to generalize the reorder to unseen pairs of blocks. Then, we identify the pairs of blocks in the source corpora (both train-ing and test) which belong to the same group. We swap them and we use the mod-ied source training corpora to realign and to build the nal translation system. We have evaluated our reordering ap-proach both in alignment and translation quality. In addition, we have used two state-of-the-art SMT systems: a Phrased-based and an Ngram-based. Experiments are reported on the EuParl task, where en-couraging results are obtained.