This article proposes the modified KNN (K Nearest Neighbor) algorithm which considers the feature similarity and is applied to the text segmentation. The words which are given as features for encoding words into numerical vectors have their own meanings and semantic relations with others, and the text segmentation is able to be viewed into a binary classification where each adjacent paragraphpair is classified into boundary or continuance. In the proposed system, a list of adjacent paragraph pairs is generated by sliding atext with the two sized window, each pair is classified by the proposed KNN version, and the boundary is put between the pairs which are classified into boundary. The proposed KNN version is empirically validated as the better approach in deciding whether each pair should be separated from each other or not in newsarticles and opinions. The significance of this research is to improve the classification performance by utilizing the feature similarities.