Direct Machine Translation Using Linguistic Similarities

Kyonghee Paik+, Hiromi Nakaiwa+, and Satoshi SHIRAI++

+ATR Spoken Language Translation Laboratories, 2-2-2, Hikaridai, Seika-cho, Soraku-gun, Kyoto 619-0288 Japan
++12-1 Ekimaehoncho, Kawasaki-ku, Kawasaki-shi, Kanagawa Pref. 210-0007 Japan
E-mail: +{kyonghee.paik,hiromi.nakaiwa}@atr.co.jp, ++shirai@nlp.ntt-at.co.jp


Abstract

Conventional approaches to machine translation require many linguistic resources either as transfer rules or parallel corpora. However, we are attempting to measure how little knowledge can be used for machine translation between similar languages, starting off with only a transfer dictionary and a target language corpus. The proposed method of machine translation exploits the linguistic similarities to achieve acceptable translation with low cost. We introduce Japanese to Korean machine translation as a case study.



Key Words

Japaese, Korean, machine translation, monolingual corpus, transfer lexicon



[ Technical Report of IEICE, pp.??-?? (August, 2003). ]