A STATISTICAL APPROACH TO VIETNAMESE-ENGLISH MACHINE TRANSLATION

Abstract

The history of Machine Translation (MT) has experienced a revolution in its methodology. That is: instead of constructing dictionaries, manual transfer rules by linguists in conventional MT systems, a new MT approach has constructed these dictionaries and transfer rules automatically by calculating probabilities on bilingual corpora. This new MT approach will statistically extract probabilistic figures of corresponding translation of words/phrases or structures as well as transpositions between two languages and appearance probabilistic of those words/phrases in a certain context. This new MT approach is the Statistical MT. In this paper, we will present the above-mentioned statistical MT approach on the Vietnamese-to-English MT system, which isn't attainable by other conventional approaches due to their requirements of morphological, syntactical and semantic analysis for Vietnamese. We have trial run this Vietnamese-English MT system on 5,000,000-word English Vietnamese bilingual corpus with initial encouraging results.