Leveraging Transliterations from Multiple Languages
Aditya Bhargava, Bradley D. Hauer, Grzegorz Kondrak · 2011
While past research on machine transliteration has focused on a single transliteration task, there exist a variety of supplemental transliterations available in other languages. Given an input for English-to-Hindi transliteration, for example, transliterations from other languages such as Japanese or Hebrew may be helpful in the transliteration process. In this paper, we propose the application of such supplemental transliterations to English-to-Hindi machine transliteration via an SVM re-ranking method with features based on n-gram alignments as well as system and alignment scores. This method achieves a relative improvement of over 10 % over the base system used on its own. We further apply this method to system combination, demonstrating just under 5 % relative improvement. 1