Selective Combination of Pivot and Direct Statistical Machine Translation Models

International Joint Conference on Natural Language Processing, pages 1174–1180, Nagoya, Japan, 14-18 October 2013. Selective Combination of Pivot and Direct Statistical Machine Translation Models Ahmed El Kholy, Nizar Habash Center for Computational Learning Systems, Columbia University {akholy,habash}@ccls.columbia.edu Gregor Leusch, Evgeny Matusov Science Applications International Corporation {gregor.leusch,evgeny.matusov}@saic.com Hassan Sawaf eBay Inc. hsawaf@ebay.com Abstract In this paper, we propose a selective com- bination approach of pivot and direct sta- tistical machine translation (SMT) models to improve translation quality. We work with Persian-Arabic SMT as a case study. We show positive results (from 0.4 to 3.1 BLEU on different direct training corpus sizes) in addition to a large reduction of pivot translation model size. 1 Introduction One of the main challenges in statistical ma- chine translation (SMT) is the scarcity of paral- lel data for many language pairs especially when the source and target languages are morpholog- ically rich. Morphological richness comes with many challenges and the severity of these chal- lenges increases when the richness and morpho- logical complexity are expressed differently in the source and target languages. A common SMT solution to the lack of par- allel data is to pivot the translation through a third language (called pivot or bridge language) for which there exist abundant parallel corpora with the source and target languages. The liter- ature covers many pivoting techniques. One of the best performing techniques, phrase pivoting (Utiyama and Isahara, 2007), builds an induced new phrase table between the source and target. One of the problems of this technique is that the size of the newly created pivot phrase table is very large (Utiyama and Isahara, 2007). Given a parallel corpus between the source and target language, combining a direct model based on this parallel corpus with a pivot model could lead to better coverage and overall translation quality. However, the combination approach needs to be optimized in order to maximize the informa- tion gain. In this paper, we propose a selective combina- tion approach of pivot and direct SMT models. The main idea is to select the relevant portions of the pivot model that do not interfere with the more trusted direct model. We show positive results for Persian-Arabic SMT (from 0.4 to 3.1 BLEU on different direct training corpus sizes). As a pos- itive side effect, we achieve a large reduction of pivot translation model size. This paper is organized as follows. Section 2 briefly discusses some related work. Section 3 presents linguistic challenges and differences be- tween Arabic and Persian. In Section 4, we dis- cuss our pivoting strategies. Then Section 5 dis- cusses our approach for selective combination. In Section 6, we present our experimental results. 2 Related Work 2.1 Pivoting Many researchers have investigated the use of piv- oting (or bridging) approaches to solve the data scarcity issue (Utiyama and Isahara, 2007; Wu and Wang, 2009; Khalilov et al., 2008; Bertoldi et al., 2008; Habash and Hu, 2009). The core idea is to introduce a pivot language, for which there exist large source-pivot …

Content truncated to 500 words