Customizing an English-Korean Machine Translation System for Patent/Technical Documents Translation
Oh‐Woog Kwon, Sung‐Kwon Choi, Kiyoung Lee, Yoon‐Hyung Roh, Young-Gil Kim · Institutional Repositories DataBase (IRDB) · 2009
This paper addresses a method for customizing an English-Korean machine translation system from general domain to patent or technical document domain.The customizing method includes the followings: (1) adapting the probabilities of POS tagger trained from general domain to the specific domain, (2) syntactically analyzing long and complex sentences by recognizing coordinate structures, and (3) selecting a proper target word using domain-specific bilingual dictionary and collocation knowledge extracted from patent or technical document corpus.The translation accuracy of the customized English-Korean patent translation system is 82.43% on the average in 5 patent categories according to the evaluation of 7 professional patent translators.The translation accuracy of the customized English-Korean technical document translation system is 81.10% and its BLEU score is 0.5185 in the evaluation test set where the average BLEU score of cross-evaluation between references is 0.6615.