Machine Translation Systems: E-K, K-E, J-K, K-J Yu Seop Kim 1 , Sung Dong Kim 1 , Seong Bae Park 2 , Jong Woo Lee 2 , Jeong Ho Chang 2 , Kyu Baek Hwang 2 , Jang Min O 2 and Yung Taek Kim 3 1 Research Institute of Advanced Computer Technology, Seoul National University, Seoul, Korea yskim, sdkim @nova.snu.ac.kr 2 School of Computer Science and Engineering, Seoul National University, Seoul, Korea sbpark, jongwoo, jhchang, mrmyself, jangmin @nova.snu.ac.kr 3 School of Computer Science and Engineering, Seoul National University, Seoul, Korea ytkim@comp.snu.ac.kr Abstract. We present four kinds of machine translation system in this description: E-K (English to Korean), K-E (Korean to English), J-K (Japanese to Korean), K-J (Korean to Japanese). Among these, E-K and K-J translation systems are published commercially, and the other systems have finished their development. This paper describes the structure and function of each system with figures and translation results. 1. E-Tran: English-Korean MT System E-Tran is an English-Korean machine translation system to translate English sentences into Korean on Microsoft Windows environment, and is composed of two versions, text version and Internet version. Internet version interacts with web browsers, such as Internet Explorer from Microsoft and Netscape Navigator, translates English sentences into Korean, and shows translated results through the browser. Figure 1 shows E-Tran Internet version with its interface and translation results. The left part of window shows the source web document and the right part shows the translated document. Text version supports the translation of user specified files. With the text version, user can translate sentences. User dictionary enables user to add specific noun or adverb words to translation system. And there are 15 special domain dictionaries; such as politics, economics, computer science and etc.. E-Tran translates documents through several steps as below. The 1 st step and the 7 th step are only needed in Internet version. The 1 st step is an HTML document pre- processing: E-Tran takes HTML documents as input and picks out sentences to be translated. In the 2 nd step, input sentences are tokenized into word stems(Lexical Analysis). In the 3 rd step, Idiomatic expressions in sentences are recognized and