Vs and OOVs: Two Problems for Translation between German and English

Proceedings of the Joint 5th Workshop on Statistical Machine Translation and MetricsMATR, pages 183–188, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics Vs and OOVs: Two Problems for Translation between German and English Sara Stymne, Maria Holmqvist, Lars Ahrenberg Link¨oping University Sweden {sarst,marho,lah}@ida.liu.se Abstract In this paper we report on experiments with three preprocessing strategies for im- proving translation output in a statistical MT system. In training, two reordering strategies were studied: (i) reorder on the basis of the alignments from Giza++, and (ii) reorder by moving all verbs to the end of segments. In translation, out-of- vocabulary words were preprocessed in a knowledge-lite fashion to identify a likely equivalent. All three strategies were im- plemented for our English↔German sys- tem submitted to the WMT10 shared task. Combining them lead to improvements in both language directions. 1 Introduction We present the Liu translation system for the con- strained condition of the WMT10 shared transla- tion task, between German and English in both di- rections. The system is based on the 2009 Liu sub- mission (Holmqvist et al., 2009), that used com- pound processing, morphological sequence mod- els, and improved alignment by reordering. This year we have focused on two issues: trans- lation of verbs, which is problematic for transla- tion between English and German since the verb placement is different with German verbs often be- ing placed at the end of sentences; and OOVs, out- of-vocabulary words, which are problematic for machine translation in general. Verb translation is targeted by trying to improve alignment, which we believe is a crucial step for verb translation since verbs that are far apart are often not aligned at all. We do this mainly by moving verbs to the end of sentences previous to alignment, which we also combine with other alignments. We trans- form OOVs into known words in a post-processing step, based on casing, stemming, and splitting of hyphenated compounds. In addition, we perform general compound splitting for German both be- fore training and translation, which also reduces the OOV rate. All results in this article are for the develop- ment test set newstest2009, on truecased output. We report Bleu scores (Papineni et al., 2002) and Meteor ranking (without WordNet) scores (Agar- wal and Lavie, 2008), using percent notation. We also used other metrics, but as they gave similar results they are not reported. For significance test- ing we used approximate randomization (Riezler and Maxwell, 2005), with p < 0.05. 2 Baseline System The 2010 Liu system is based on the PBSMT base- line system for the WMT shared translation task1. We use the Moses toolkit (Koehn et al., 2007) for decoding and to train translation models, Giza++ (Och and Ney, 2003) for word alignment, and the SRILM toolkit (Stolcke, 2002) to train language models. The main difference to the WMT base- line is that the Liu system is trained on truecased data, as in Koehn et al. (2008), instead of lower- cased data. This means that there is no …

Content truncated to 500 words