MaTrEx: The DCU MT System for WMT 2008

Proceedings of the Third Workshop on Statistical Machine Translation, pages 171–174, Columbus, Ohio, USA, June 2008. c⃝2008 Association for Computational Linguistics MATREX: the DCU MT System for WMT 2008 John Tinsley, Yanjun Ma, Sylwia Ozdowska, Andy Way National Centre for Language Technology Dublin City University Dublin 9, Ireland {jtinsley, yma, sozdowska, away}@computing.dcu.ie Abstract In this paper, we give a description of the ma- chine translation system developed at DCU that was used for our participation in the eval- uation campaign of the Third Workshop on Statistical Machine Translation at ACL 2008. We describe the modular design of our data- driven MT system with particular focus on the components used in this participation. We also describe some of the significant modules which were unused in this task. We participated in the EuroParl task for the following translation directions: Spanish– English and French–English, in which we em- ployed our hybrid EBMT-SMT architecture to translate. We also participated in the Czech– English News and News Commentary tasks which represented a previously untested lan- guage pair for our system. We report results on the provided development and test sets. 1 Introduction In this paper, we present the Data-Driven MT sys- tems developed at DCU, MATREX (Machine Trans- lation using Examples). This system is a hybrid sys- tem which exploits EBMT and SMT techniques to build a combined translation model. We participated in both the French–English and Spanish–English EuroParl tasks. In these two tasks, we monolingually chunk both source and target sides of the dataset using a marker-based chunker (Gough and Way, 2004). We then align these chunks using a dynamic programming, edit-distance-style algorithm and combine them with phrase-based SMT-style chunks into a single translation model. We also participated in the Czech–English News Commentary and News tasks. This language pair represents a new challenge for our system and pro- vides a good test of its flexibility. The remainder of this paper is organised as fol- lows: Section 2 details the various components of our system, in particular the chunking and chunk alignment strategies used for the shared task. In Sec- tion 3, we outline the complete system setup for the shared task, and in Section 4 we give some results and discussion thereof. 2 The MATREX System The MATREX system is a modular hybrid data- driven MT system, built following established De- sign Patterns, which exploits aspects of both the EBMT and SMT paradigms. It consists of a num- ber of extendible and re-implementable modules, the most significant of which are: • Word Alignment Module: outputs a set of word alignments given a parallel corpus, • Chunking Module: outputs a set of chunks given an input corpus, • Chunk Alignment Module: outputs aligned chunk pairs given source and target chunks ex- tracted from comparable corpora, • Decoder: returns optimal translation given a set of aligned sentence, chunk/phrase and word pairs. In some cases, these modules may comprise wrappers around pre-existing software. For exam- ple, our system configuration for the shared task incorporates a wrapper around …

Content truncated to 500 words