We address the problem of transliteratingEnglish names using Chinese orthographyin support of cross-lingual speech andtext processing applications. We demonstratethe application of statistical machinetranslation techniques to "translate"the phonemic representation of an Englishname, obtained by using an automatictext-to-speech system, to a sequenceof initials and finals, commonly used subwordunits of pronunciation for Chinese.We then use another statistical translationmodel to map the initial/final sequenceto Chinese characters. We also presentan evaluation of this module in retrievalof Mandarin spoken documents from theTDT corpus using English text queries.
展开▼