The NTT Statistical Machine Translation System for IWSLT2005

The NTT Statistical Machine Translation System for IWSLT2005 Hajime Tsukada, Taro Watanabe, Jun Suzuki, Hideto Kazawa, and Hideki Isozaki NTT Communication Science Laboratories {tsukada,taro,jun,kazawa,isozaki}@cslab.kecl.ntt.co.jp Abstract This paper reports the NTT statistical translation system par- ticipating in the evaluation campaign of IWSLT 2005. The NTT system is based on a phrase translation model and uti- lizes a large number of features with a log-linear model. We studied the various features recently developed in this re- search field and evaluate the system using supplied data as well as publicly available Chinese, Japanese, and English data. Despite domain mismatch, additional data helped im- prove translation accuracy. 1. Introduction Recently, phrase-based translation combined with other fea- tures by log-linear models has become the standard technique for statistical machine translation. Shared task based work- shops of machine translation including IWSLT and NIST Machine Translation Workshops showed which features ef- fectively improve translation accuracy. However, it remains unclear whether using of these features all together with our system is helpful. One unavoidable problem with statistical approaches is training data preparation. Since the amount of training data is generally limited, how to utilize similar monolingual or bilingual resources is an important research topic in statisti- cal machine translation. In this evaluation campaign, we studied the use of a large number of reportedly effective features with our system and also evaluated both additional monolingual and bilingual cor- pus to improve translation accuracies. 2. Log-linear Models Our system adopts the following log-linear decision rule to obtain the maximum likely translation: ˆeI 1 = argmax eI 1 1 Z(f J 1 ) exp  � j λjfj(eI 1, f J 1 )   , where fj(eI 1, f J 1 ) represents a feature function and Z(f J 1 ) denotes a normalization term. Feature function scaling fac- tors λj are efficiently computed based either on the maxi- mum likelihood criterion [1] or the minimum error rate crite- rion [2]. Our system adopts the latter criterion in the experi- ments. One advantage of log-linear models is the ability to eas- ily combine various features relating to translation models, language models, and lexical reordering models. The feature details are described in the following sections. 3. Features 3.1. Language Model Features In statistical machine translation, improving language mod- els strongly impacts translation accuracy. Especially re- cently, the power of long-span n-grams and the use of huge amounts of training data have been reported [3]. In this eval- uation campaign, we combined several long n-gram language models as features. To train language models from various corpora in differ- ent domains, corpus weighting is necessary to fit the trained language models to the test set domain. Log-linear models naturally provide weighting of a language model trained by each corpus. We used the following long n-gram language models: • 6-gram • Class-based 9-gram • Prefix-4 9-gram • Suffix-4 9-gram All n-grams are based on mixed casing. The prefix-4 (suffix- 4) language model takes only 4-letter prefixes (suffixes) of English words. Prefix-4 (suffix-4) roughly means …

Content truncated to 500 words