In the Japanese language, as a predicate is placed at the end of a sentence, the content of a sentence cannot be inferred until reaching the end. However, when the content is complicated and the sentence is long, people want to know at an earlier stage in the sentence whether the content is negative, affirmative, or interrogative. In Japanese, the grammatical form called the KO-OU relation exists. The KO-OU relation is a kind of concord. If a KO element appears, then an OU element appears in the latter part of a sentence. It is being pointed out that the KO-OU relation gives advance notice to the element that appears in the latter part of a sentence. In this paper, we present the method of extracting automatically the KO-OU expression data from large-scale electronic corpus and verify the usefulness of the KO-OU expression data.