K Nearest Neighbor for Text Summarization using Feature Similarity

Abstract

In this research, we propose a particular version of KNN (K Nearest Neighbor) where the similarity between feature vectors is computed considering the similarity among attributes or features as well as one among values. The task of text summarization is viewed into the binary classification task where each paragraph or sentence is classified into the essence or non-essence, and in previous works, improved results are obtained by the proposed version in the text classification and clustering. In this research, we define the similarity which considers both attributes and attribute values, modify the KNN into the version based on the similarity, and use the modified version as the approach to the text summarization task. As the benefits from this research, we may expect the more compact representation of data items and the better performance. Therefore, the goal of this research is to implement the text summarization algorithm which represents data items more compactly and provides the more reliability.