Automatic Extraction of Translational Japanese-KATAKANA and English Word Pairs

The method to automatically extract translational Japanese-KATAKANA and English word pairs from bilingual corpora is proposed. The method applies all the existing transliteration rules to each mora unit in a KATAKANA word, and extract English word which matched or partially-matched to one of these transliteration candidates as translation. For instance, if there is a word `Ot ' (graph) in Japanese part of bilingual corpora, we generate such transliteration candidates as <graf>, <graph>, <gulerph>,... and extract similar words from English part of corpora. The method worked fairly well, achieving 83-100% precision at 75% recall against eight corpora we used for evaluation.

• Type: Journal
• Year: 2002
• Where: IJCPOL
• Authors: Keita Tsuji
• Keywords: Bilingual Corpora, English Word, IJCPOL 2002, Transliteration Candidates