Refinements in BTG-based Statistical Machine Translation

Refinements in BTG-based Statistical Machine Translation Deyi Xiong, Min Zhang, Aiti Aw Human Language Technology Institute for Infocomm Research 21 Heng Mui Keng Terrace Singapore 119613 {dyxiong, mzhang, aaiti}@i2r.a-star.edu.sg Haitao Mi, Qun Liu and Shouxun Lin Key Lab of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences Beijing China, 100080 {htmi, liuqun, sxlin}@ict.ac.cn Abstract Bracketing Transduction Grammar (BTG) has been well studied and used in statistical machine translation (SMT) with promising results. However, there are two major issues for BTG-based SMT. First, there is no effec- tive mechanism available for predicting or- ders between neighboring blocks in the orig- inal BTG. Second, the computational cost is high. In this paper, we introduce two re- finements for BTG-based SMT to achieve better reordering and higher-speed decod- ing, which include (1) reordering heuristics to prevent incorrect swapping and reduce search space, and (2) special phrases with tags to indicate sentence beginning and end- ing. The two refinements are integrated into a well-established BTG-based Chinese-to- English SMT system that is trained on large- scale parallel data. Experimental results on the NIST MT-05 task show that the proposed refinements contribute significant improve- ment of 2% in BLEU score over the baseline system. 1 Introduction Bracket transduction grammar was proposed by Wu (1995) and firstly employed in statistical machine translation in (Wu, 1996). Because of its good trade- off between efficiency and expressiveness, BTG re- striction is widely used for reordering in SMT (Zens et al., 2004). However, BTG restriction does not provide a mechanism to predict final orders between two neighboring blocks. To solve this problem, Xiong et al. (2006) proposed an enhanced BTG with a maximum en- tropy (MaxEnt) based reordering model (MEBTG). MEBTG uses boundary words of bilingual phrases as features to predict their orders. Xiong et al. (2006) reported significant performance im- provement on Chinese-English translation tasks in two different domains when compared with both Pharaoh (Koehn, 2004) and the original BTG us- ing flat reordering. However, error analysis of the translation output of Xiong et al. (2006) reveals that boundary words predict wrong swapping, espe- cially for long phrases although the MaxEnt-based reordering model shows better performance than baseline reordering models. Another big problem with BTG-based SMT is the high computational cost. Huang et al. (2005) re- ported that the time complexity of BTG decoding with m-gram language model is O(n3+4(m−1)). If a 4-gram language model is used (common in many current SMT systems), the time complexity is as high as O(n15). Therefore with this time complexity translating long sentences is time-consuming even with highly stringent pruning strategy. To speed up BTG decoding, Huang et al. (2005) adapted the hook trick which changes the time complexity from O(n3+4(m−1)) to O(n3+3(m−1)). However, the implementation of the hook trick with pruning is quite complicated. Another method to in- crease decoding speed is cube pruning proposed by Chiang (2007) which reduces search space signifi- cantly. In this paper, we propose two refinements to ad- dress the two issues, including (1) …

Content truncated to 500 words