The RWTH System Combination System for WMT 2010

Proceedings of the Joint 5th Workshop on Statistical Machine Translation and MetricsMATR, pages 315–320, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics The RWTH System Combination System for WMT 2010 Gregor Leusch and Hermann Ney RWTH Aachen University Aachen, Germany {leusch,ney}@cs.rwth-aachen.de Abstract RWTH participated in the System Combi- nation task of the Fifth Workshop on Sta- tistical Machine Translation (WMT 2010). For 7 of the 8 language pairs, we com- bine 5 to 13 systems into a single con- sensus translation, using additional n-best reranking techniques in two of these lan- guage pairs. Depending on the language pair, improvements versus the best sin- gle system are in the range of +0.5 and +1.7 on BLEU, and between −0.4 and −2.3 on TER. Novel techniques compared with RWTH’s submission to WMT 2009 include the utilization of n-best reranking techniques, a consensus true casing ap- proach, a different tuning algorithm, and the separate selection of input systems for CN construction, primary/skeleton hy- potheses, HypLM, and true casing. 1 Introduction The RWTH approach to MT system combination is a refined version of the ROVER approach in ASR (Fiscus, 1997), with additional steps to cope with reordering between different hypotheses, and to use true casing information from the input hy- potheses. The basic concept of the approach has been described by Matusov et al. (2006). Several improvements have been added later (Matusov et al., 2008). This approach includes an enhanced alignment and reordering framework. In con- trast to existing approaches (Jayaraman and Lavie, 2005; Rosti et al., 2007), the context of the whole corpus rather than a single sentence is considered in this iterative, unsupervised procedure, yielding a more reliable alignment. Majority voting on the generated lattice is performed using prior weights for each system as well as other statistical mod- els such as a special n-gram language model. In addition to lattice rescoring, n-best list reranking techniques can be applied to n best paths of this lattice. True casing is considered a separate step in RWTH’s approach, which also takes the input hypotheses into account. The pipeline, and consequently the description of the pipeline given in this paper, is based on our pipeline for WMT 2009 (Leusch et al., 2009), with several extensions as described. 2 System Combination Algorithm In this section we present the details of our system combination method. Figure 1 gives an overview of the system combination architecture described in this section. After preprocessing the MT hy- potheses, pairwise alignments between the hy- potheses are calculated. The hypotheses are then reordered to match the word order of a selected primary or skeleton hypothesis. From this, we create a lattice which we then rescore using sys- tem prior weights and a language model (LM). The single best path in this CN then constitutes the consensus translation; alternatively the n best paths are generated and reranked using additional statistical models. The consensus translation is then true cased and postprocessed. 2.1 Word Alignment The proposed alignment approach is a statistical one. …

Content truncated to 500 words