Using Syntactic Head Information in Hierarchical Phrase-Based Translation

Proceedings of the 7th Workshop on Statistical Machine Translation, pages 232–242, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics Using Syntactic Head Information in Hierarchical Phrase-Based Translation Junhui Li Zhaopeng Tu† Guodong Zhou‡ Josef van Genabith Centre for Next Generation Localisation School of Computing, Dublin City University † Key Lab. of Intelligent Info. Processing Institute of Computing Technology, Chinese Academy of Sciences ‡School of Computer Science and Technology Soochow University, China {jli,josef}@computing.dcu.ie tuzhaopeng@ict.ac.cn gdzhou@suda.edu.cn Abstract Chiang’s hierarchical phrase-based (HPB) translation model advances the state-of-the-art in statistical machine translation by expanding conventional phrases to hierarchical phrases – phrases that contain sub-phrases. How- ever, the original HPB model is prone to over- generation due to lack of linguistic knowl- edge: the grammar may suggest more deriva- tions than appropriate, many of which may lead to ungrammatical translations. On the other hand, limitations of glue grammar rules in the original HPB model may actually pre- vent systems from considering some reason- able derivations. This paper presents a sim- ple but effective translation model, called the Head-Driven HPB (HD-HPB) model, which incorporates head information in translation rules to better capture syntax-driven informa- tion in a derivation. In addition, unlike the original glue rules, the HD-HPB model allows improved reordering between any two neigh- boring non-terminals to explore a larger re- ordering search space. An extensive set of ex- periments on Chinese-English translation on four NIST MT test sets, using both a small and a large training set, show that our HD- HPB model consistently and statistically sig- nificantly outperforms Chiang’s model as well as a source side SAMT-style model. 1 Introduction Chiang’s hierarchical phrase-based (HPB) transla- tion model utilizes synchronous context free gram- mar (SCFG) for translation derivation (Chiang, 2005; Chiang, 2007) and has been widely adopted in statistical machine translation (SMT). Typically, such models define two types of translation rules: hierarchical (translation) rules which consist of both terminals and non-terminals, and glue (grammar) rules which combine translated phrases in a mono- tone fashion. However, due to lack of linguistic knowledge, Chiang’s HPB model contains only one type of non-terminal symbol X, often making it difficult to select the most appropriate translation rules.1 One important research question is therefore how to refine the non-terminal category X using linguis- tically motivated information: Zollmann and Venu- gopal (2006) (SAMT) e.g. use (partial) syntactic categories derived from CFG trees while Zollmann and Vogel (2011) use word tags, generated by ei- ther POS analysis or unsupervised word class in- duction. Almaghout et al. (2011) employ CCG- based supertags. Mylonakis and Sima’an (2011) use linguistic information of various granularities such as Phrase-Pair, Constituent, Concatenation of Con- stituents, and Partial Constituents, where applica- ble. By contrast, and inspired by previous work in parsing (Charniak, 2000; Collins, 2003), our Head- Driven HPB (HD-HPB) model is based on the in- tuition that linguistic heads provide important in- formation about a constituent or distributionally de- fined fragment, as in HPB. We identify heads using linguistically motivated dependency parsing, and use head …

Content truncated to 500 words