The RWTH Phrase-based Statistical Machine Translation System

The RWTH Phrase-based Statistical Machine Translation System Richard Zens, Oliver Bender, Saˇsa Hasan, Shahram Khadivi, Evgeny Matusov, Jia Xu, Yuqi Zhang, Hermann Ney Human Language Technology and Pattern Recognition Lehrstuhl f¨ur Informatik VI, Computer Science Department RWTH Aachen University, D-52056 Aachen, Germany {zens,bender,hasan,khadivi,matusov,xujia,yzhang,ney}@cs.rwth-aachen.de Abstract We give an overview of the RWTH phrase-based statistical machine translation system that was used in the evaluation campaign of the International Workshop on Spoken Lan- guage Translation 2005. We use a two pass approach. In the first pass, we gen- erate a list of the N best translation candidates. The second pass consists of rescoring and reranking this N-best list. We will give a description of the search algorithm as well as the models that are used in each pass. We participated in the supplied data tracks for man- ual transcriptions for the following translation direc- tions: Arabic-English, Chinese-English, English-Chinese and Japanese-English. For Japanese-English, we also partic- ipated in the C-Star track. In addition, we performed trans- lations of automatic speech recognition output for Chinese- English and Japanese-English. For both language pairs, we translated the single-best ASR hypotheses. Additionally, we translated Chinese ASR lattices. 1. Introduction We give an overview of the RWTH phrase-based statistical machine translation system that was used in the evaluation campaign of the International Workshop on Spoken Lan- guage Translation (IWSLT) 2005. We use a two pass approach. First, we generate a word graph and extract a list of the N best translation candidates. Then, we apply additional models in a rescoring/reranking approach. This work is structured as follows: first, we will review the statistical approach to machine translation and introduce the notation that we will use in the later sections. Then, we will describe the models and algorithms that are used for gen- erating the N-best lists, i.e., the first pass. In Section 4, we will describe the models that are used to rescore and rerank this N-best list, i.e., the second pass. Afterward, we will give an overview of the tasks and discuss the experimental results. 1.1. Source-channel approach to SMT In statistical machine translation, we are given a source lan- guage sentence f J 1 = f1 . . . fj . . . fJ, which is to be trans- lated into a target language sentence eI 1 = e1 . . . ei . . . eI. Among all possible target language sentences, we will choose the sentence with the highest probability: ˆe 1ˆI = argmax I,eI 1 � Pr(eI 1|f J 1 ) � (1) = argmax I,eI 1 � Pr(eI 1) · Pr(f J 1 |eI 1) � (2) This decomposition into two knowledge sources is known as the source-channel approach to statistical machine transla- tion [1]. It allows an independent modeling of the target lan- guage model Pr(eI 1) and the translation model Pr(f J 1 |eI 1)1. The target language model describes the well-formedness of the target language sentence. The translation model links the source language sentence to the target language sentence. The …

Content truncated to 500 words