Direct Combination of Spelling and Pronunciation Information for Robust Back-Transliteration

Transliterating words and names from one language to another is a frequent and highly productive phenomenon. For example, English word cache is transliterated in Japanese as キャツシェ “kyasshu”. Transliteration is information losing since important distinctions are not always preserved in the process. Hence, automatically converting transliterated words back into their original form is a real challenge. Nonetheless, due to its wide applicability in MT and CLIR, it is an interesting problem from a practical point of view. In this paper, we demonstrate that back-transliteration accuracy can be improved by directly combining grapheme-based (i.e. spelling) and phoneme-based (i.e. pronunciation) information. Rather than producing back-transliterations based on grapheme and phoneme model independently and then interpolating the results, we propose a method of first combining the sets of allowed rewrites (i.e. edits) and then calculating the back-transliterations using the combined set. Evaluation on both Japanese and Chinese transliterations shows that direct combination increases robustness and positively affects back-transliteration accuracy.