The practical English-Korean machine translation system must have an ability of the fast parsing and the accurate target sentence generation for the long sentences. The parsing complexity exponentially increases as the length of the input sentences increases. It is also difficult to generate accurate translation for the long sentences. This paper proposes the intra-sentence segmentation method which uses the finite automata to identify the candidate segmentation positions in the input sentences. We define the context of the segmentable positions using the word types or the part-of-speeches of the words. Then, we construct the finite automata based on the defined context. We recognize the segmentable positions using the constructed finite automata and collect the candidate segmentation positions. In case of multiple candidate segmentation positions, we determine the segmentation positions by considering the priority among the word types of the candidate segmentation positions and the positions in the input sentence. In the experiment, we compare the segmentation precision and the recall with those by other methods. Through the comparison, we show that the proposed intra-sentence segmentation method improves the segmentation performance.