Learning Improved Reordering Models for Urdu, Farsi and Italian using SMT
Rohit Gupta, Raj Nath Patel, Ritesh Shah · International Conference on Computational Linguistics · 2012
This paper presents some experiments which have been carried out as part of a shared task for the workshop “Reordering for Statistical Machine Translation” (RSMT, collocated with COLING 2012). The shared task objective is to learn reordering models by making use of a manually word-aligned, bilingual parallel data. We view this task as that of a statistical machine translation (SMT) system which implicitly employ such models. These models are obtained using empirical methods and machine learning techniques. We have therefore used “Moses”; a state of the art SMT system to conduct experiments for the task at hand. The training and the development datasets used for the experiments have been provided by RSMT and we report our work on three pair of languages namely English-Urdu, English- Farsi and English- Italian.