Hierarchical System Combination for Machine Translation

Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pp. 277–286, Prague, June 2007. c⃝2007 Association for Computational Linguistics Hierarchical System Combination for Machine Translation Fei Huang IBM T.J. Watson Research Center Yorktown Heights, NY 10562 huangfe@us.ibm.com Kishore Papineni ∗ Yahoo! Research New York, NY 10011 kpapi@yahoo-inc.com Abstract Given multiple translations of the same source sentence, how to combine them to produce a translation that is better than any single system output? We propose a hier- archical system combination framework for machine translation. This framework inte- grates multiple MT systems’ output at the word-, phrase- and sentence- levels. By boosting common word and phrase trans- lation pairs, pruning unused phrases, and exploring decoding paths adopted by other MT systems, this framework achieves bet- ter translation quality with much less re- decoding time. The full sentence translation hypotheses from multiple systems are addi- tionally selected based on N-gram language models trained on word/word-POS mixed stream, which further improves the transla- tion quality. We consistently observed sig- nificant improvements on several test sets in multiple languages covering different gen- res. 1 Introduction Many machine translation (MT) frameworks have been developed, including rule-based transfer MT, corpus-based MT (statistical MT and example-based MT), syntax-based MT and the hybrid, statistical MT augmented with syntactic structures. Different MT paradigms have their strengths and weaknesses. ∗This work was done when the author was at IBM Research. Systems adopting the same framework usually pro- duce different translations for the same input, due to their differences in training data, preprocessing, alignment and decoding strategies. It is beneficial to design a framework that combines the decoding strategies of multiple systems as well as their out- puts and produces translations better than any single system output. More recently, within the GALE1 project, multiple MT systems have been developed in each consortium, thus system combination be- comes more important. Traditionally, system combination has been con- ducted in two ways: glass-box combination and black-box combination. In the glass-box combi- nation, each MT system provides detailed decod- ing information, such as word and phrase transla- tion pairs and decoding lattices. For example, in the multi-engine machine translation system (Nirenburg and Frederking, 1994), target language phrases from each system and their corresponding source phrases are recorded in a chart structure, together with their confidence scores. A chart-walk algorithm is used to select the best translation from the chart. To com- bine words and phrases from multiple systems, it is preferable that all the systems adopt similar prepro- cessing strategies. In the black-box combination, individual MT sys- tems only output their top-N translation hypothe- ses without decoding details. This is particularly appealing when combining the translation outputs from COTS MT systems. The final translation may be selected by voted language models and appropri- ate confidence rescaling schemes ((Tidhar and Kuss- 1http://www.darpa.mil/ipto/programs/gale/index.htm 277 ner, 2000) and (Nomoto, 2004)). (Mellebeek et al., 2006) decomposes source sentences into meaning- ful constituents, translates them with component MT systems, then selects the …

Content truncated to 500 words