Improving alignment for SMT by reordering and augmenting the training corpus
2009 (English)In: Proceedings of the Fourth Workshop on Statistical Machine Translation (WMT09), Athens, Greece, 2009, 120-124 p.Conference paper (Refereed)
We describe the LIU systems for English-German and German-English translation in the WMT09 shared task. We focus on two methods to improve the word alignment: (i) by applying Giza++ in a second phase to a reordered training corpus, where reordering is based on the alignments from the first phase, and (ii) by adding lexical data obtained as high-precision alignments from a different word aligner. These methods were studied in the context of a system that uses compound processing, a morphological sequence model for German, and a part-of-speech sequence model for English. Both methods gave some improvements to translation quality as measured by Bleu and Meteor scores, though not consistently. All systems used both out-of-domain and in-domain data as the mixed corpus had better scores in the baseline configuration.
Place, publisher, year, edition, pages
Athens, Greece, 2009. 120-124 p.
Machine translation, reordering, word alignment
National CategoryLanguage Technology (Computational Linguistics) Computer Science
IdentifiersURN: urn:nbn:se:liu:diva-58978OAI: oai:DiVA.org:liu-58978DiVA: diva2:347917
The Fourth Workshop on Statistical Machine Translation (WMT09)