Translating Common English and Chinese Verb-Noun Pairs in Technical Documents with Collocational and Bilingual Information

Translating Common English and Chinese Verb-Noun Pairs in Technical Documents with Collocational and Bilingual Information Yi-Hsuan Chuang† Chao-Lin Liu§ Jing-Shin Chang↑ †§ National Chengchi University, Taiwan; ↑National Chi Nan University, Taiwan {†g9804, §chaolin}@cs.nccu.edu.tw, ↑jshin@csie.ncnu.edu.tw Abstract. We studied a special case for the translation of English verbs in verb-object pairs. Researchers have studied the effects of the linguistic information about the verbs being translated, and many have reported how considering the objects of the verbs will facilitate the quality of translations. In this study, we took an extreme venue – assuming the availa- bility of the Chinese translations of the English objects. We explored the issue with thou- sands of samples that we extracted from 2011 NTCIR PatentMT workshop. The results indicated that, when the English verbs and objects were known, the information about the object’s Chinese translation could still improve the quality of the verb’s translations but not quite significantly. Keywords: Machine Translation, Feature Comparison 1 Introduction Researchers have studied extensively the problems related to verbs (e.g., Dorr et al., 2002; La- pata and Brew, 2004) and phrases-based translations (e.g., Chuang et al., 2005; Koehn et al., 2003). Some techniques were developed for text of special domains (Seneff et al., 2006). The techniques are applicable in many real-world problems, including computer-assisted language learning (Chang et al., 2008) and cross-language information retrieval (Chen et al., 2000). We work on the processing of patent documents (Lu et al., 2010; Yokoama and Okuyama, 2009), and present an experience in translating common verbs and their direct object based on bilingual contextual information. In this study, we took an extreme assumption of the availabili- ty of the Chinese translations of the English objects to examine whether the extra information will improve the quality of verbs’ translation. The proposed methods are special in that we are crossing the boundary between translation models and language models, by considering infor- mation of the target language in the translation task. The purpose of conducting such experi- ments was to investigate how the availability of such bilingual and collocational information might contribute to the translation quality. It is understood and expected by many that the Chi- nese translations of English objects might not be available for all cases and that such informa- tion should be applied with many other features to achieve high translation quality. The experiments were conducted with the training data available to the participants of the 2011 NTCIR Patent MT task. The corpus contains one million pairs of Chinese and English pairs. We explored four different methods to determine the verb’s Chinese translation. These methods utilized the bilingual and contextual information about the English verb in different ways. Effects of these methods were compared based on experimental evaluation which con- ducted with 35 thousands of verb-object pairs extracted from the NTCIR corpus. (Since objects are nouns, we will refer to verb-object pairs as verb-noun pairs or VN pairs to simplify the wording.) We provide a broad outline of our work in Section 2, present our methods for …

Content truncated to 500 words