Research for Tibetan-Chinese Name Transliteration Based on Multi-granularity

摘要

In order to solve the problem of data sparseness caused by less training corpus in Tibetan-Chinese transliteration, this paper analyzes the alignment granularity of Tibetan-Chinese names as the research object and uses the pronunciation feature to reduce the corresponding relationships. The method of transliteration of Tibetan and Chinese names and the design of related experiments is comparable with traditional methods and improve the top-1 accuracy of transliteration of Tibetan and Chinese names to 65.72%. The experimental results show that the method can improve the accuracy of Tibetan-Chinese name transliteration.