An Ensemble of Grapheme and Phoneme for Machine Transliteration

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 450 – 461, 2005. © Springer-Verlag Berlin Heidelberg 2005 An Ensemble of Grapheme and Phoneme for Machine Transliteration Jong-Hoon Oh and Key-Sun Choi Department of Computer Science, KAIST/KORTERM/BOLA, 373-1 Guseong-dong, Yuseong-gu, Daejeon, 305-701, Republic of Korea {rovellia, kschoi}@world.kaist.ac.kr Abstract. Machine transliteration is an automatic method to generate characters or words in one alphabetical system for the corresponding characters in another alphabetical system. There has been increasing concern on machine translitera- tion as an assistant of machine translation and information retrieval. Three ma- chine transliteration models, including “grapheme-based model”, “phoneme- based model”, and “hybrid model”, have been proposed. However, there are few works trying to make use of correspondence between source grapheme and phoneme, although the correspondence plays an important role in machine transliteration. Furthermore there are few works, which dynamically handle source grapheme and phoneme. In this paper, we propose a new transliteration model based on an ensemble of grapheme and phoneme. Our model makes use of the correspondence and dynamically uses source grapheme and phoneme. Our method shows better performance than the previous works about 15~23% in English-to-Korean transliteration and about 15~43% in English-to-Japanese transliteration. 1 Introduction Machine transliteration is an automatic method to generate characters or words in one alphabetical system for the corresponding characters in another alphabetical system. For example, English word data is transliterated into Korean ‘deita’ 1 and Japanese ‘deeta’. Transliteration is used to phonetically translate proper names and technical terms especially from languages in Roman alphabets to languages in non-Roman alphabets such as from English to Korean, Japanese, and Chinese and so on. There has been increasing concern on machine transliteration as an assistant of Machine Translation (MT) [2], [10], mono-lingual information retrieval (MLIR) [8], [11] and cross-lingual information retrieval (CLIR) [6]. In the area of MLIR and CLIR, ma- chine transliteration bridges the gap between a transliterated localized form and its original form by generating all possible transliterated forms from each original form. Especially for CLIR, machine transliteration gives a help to query translation where proper names and technical terms frequently appear in source language queries. In the area of MT, machine transliteration prevents translation failure when translations of 1 In this paper, target language transliterations are represented with their Romanization form in a quotation mark (‘’) . An Ensemble of Grapheme and Phoneme for Machine Transliteration 451 proper names and technical terms are not registered in a translation dictionary. A machine transliteration system, therefore, may affect the performance of MT, MLIR, and CLIR system. Three machine transliteration models have been studied: called “grapheme2-based transliteration model (ψG)” [7], [8], [9], [11], [12], [13], “phoneme3-based translit- eration model (ψP)” [10], [12], and “hybrid transliteration model (ψH)” [2], [4], [12]. ψG and ψP are classified in terms of units to be transliterated. ψG is referred to the direct model because it directly transforms source language graphemes to target language graphemes without any phonetic knowledge of source language words. ψP is called the pivot model because …

Content truncated to 500 words