Direct Translation Model 2

Proceedings of NAACL HLT 2007, pages 57–64, Rochester, NY, April 2007. c⃝2007 Association for Computational Linguistics Direct Translation Model 2 Abraham Ittycheriah and Salim Roukos IBM T.J. Watson Research Center 1101 Kitchawan Road Yorktown Heights, NY 10598 {abei,roukos}@us.ibm.com Abstract This paper presents a maximum entropy ma- chine translation system using a minimal set of translation blocks (phrase-pairs). While recent phrase-based statistical machine trans- lation (SMT) systems achieve significant im- provement over the original source-channel sta- tistical translation models, they 1) use a large inventory of blocks which have significant over- lap and 2) limit the use of training to just a few parameters (on the order of ten). In con- trast, we show that our proposed minimalist system (DTM2) achieves equal or better per- formance by 1) recasting the translation prob- lem in the traditional statistical modeling ap- proach using blocks with no overlap and 2) re- lying on training most system parameters (on the order of millions or larger). The new model is a direct translation model (DTM) formu- lation which allows easy integration of addi- tional/alternative views of both source and tar- get sentences such as segmentation for a source language such as Arabic, part-of-speech of both source and target, etc. We show improvements over a state-of-the-art phrase-based decoder in Arabic-English translation. 1 Introduction Statistical machine translation takes a source se- quence, S = [s1 s2 . . . sK], and generates a target sequence, T ∗ = [t1 t2 . . . tL], by finding the most likely translation given by: T ∗ = arg max T p(T |S). 1.1 Block selection Recent statistical machine translation (SMT) al- gorithms generate such a translation by incorpo- rating an inventory of bilingual phrases (Och and Ney, 2000). A m-n phrase-pair, or block, is a se- quence of m source words paired with a sequence of n target words. The inventory of blocks in cur- rent systems is highly redundant. We illustrate the redundancy using the example in Table 1 which lljnp Almrkzyp llHzb Al$ywEy AlSyny the Politburo of the Central Committee of the Chinese Communist Party Almktb AlsyAsy Figure 1: Example of Arabic snipet and alignment to its English translation. shows a set of phrases that cover the two-word Arabic fragment “lljnp Almrkzyp” whose align- ment and translation is shown in Figure 1. One notices the significant overlap between the vari- ous blocks including the fact the output target se- quence “of the central committee” can be pro- duced in at least two different ways: 1) as 2-4 block “lljnp Almrkzyp | of the central committee” cov- ering the two Arabic words, or 2) by using the 1- 3 block “Almrkzyp | of the central” followed by covering the first Arabic word with the 1-1 block “lljnp | committee”. In addition, if one adds one more word to the Arabic fragment in the third posi- tion such as the block “AlSyny | chinese” the over- lap increases significantly and more alternate possi- bilities are available to produce an output …

Content truncated to 500 words