Translation divergence in English-Hindi MT

Translation Divergence in English-Hindi MT R. Mahesh K. Sinha and Anil Thakur Department of Computer Science and Engineering Indian Institute of Technology, Kanpur 208016 INDIA E-mail: {rmk, anilt}@iitk.ac.in Abstract. Divergence related to mapping patterns between two or more natural languages is a common phenomenon. The patterns of divergence between two languages need to be identified and strategies devised to handle them to obtain correct translation from one language to another. In the literature on MT, some attempts have been made to classify the types of translation divergence between a pair of natural languages. However, the issue of linguistic divergence is such a complex phenomenon that a lot more need to be done in this area to identify further classes of divergence, their implications and inter-relatedness as well as the approaches to handle them. In this paper, we take Dorr’s (1994) classification of translation divergence as base and examine the translation patterns between Hindi and English to locate further details and implications of these divergences. We attempt to identify the potential topics that fall under divergence and cannot directly or indirectly be accounted for or accommodated within the existing classification. Our primary goal is to identify different patterns of translation divergence from Hindi to English and the vice versa and, on the basis of that, suggest an augmentation in the classification of translation divergence. 1. Introduction Hindi-English language pair for (machine) translation presents a rich case of divergence at different grammatical as well as extra- grammatical levels. It is important to identify the different types of divergences to obtain correct translation for Hindi sentences to English and the vice-versa. The translation divergences has been examined in the literature on MT from different theoretical perspectives for the purpose of their proper classification and handling (Dorr 1990a, 1990b, 1993, 1994, Barnett et al 1991a, Barnett et al 1991b, see Dorr 1994 for a brief review of them). The issue of translation divergence has not been discussed in detail in the context of machine translation between English and Hindi. However, the issue has recently caught attention of scholars working in the area of machine translation in English and Hindi (including other Indian languages) but the existing works (Dave et al, 2001; Gupta et al, 2003; Goyal et al, 2004; Sinha and Thakur, 2004) are only a beginning in the vast and complex area of translation divergence. Dave et al (2001) discuss some of the major classes of translation divergence as proposed in Dorr (1993) and oultline a UNL- based interlingua approach for the handling of some of the translation divergences between English and Hindi. Gupta et al (2003) discuss some of the translation divergences for English- Hindi MT, based on the classification proposed in Dorr (1994) and suggest a unified approach for their identification and resolution. However, the issue of translation divergence is complex one and a number of significant divergence issues have remained out of the scope in the existing works on the topic. We also notice that the classification of translation divergence …

Content truncated to 500 words