Long Text Segmentation by String Vector based KNN

Abstract

In this research, we propose the string vector based version of KNN as the approach to the text segmentation. The text segmentation may be interpreted into the text classification, and encoding texts into string vectors improved previously the text classification performance. In this research, we encode sentence pairs or paragraph pairs into string vectors, and apply the string vector based version of KNN to the classification task mapped from the text segmentation. As the benefits from this research, we expect the better performance, the more compact representation, and the more transparency by doing so. The goal of this research is to improve the performance of the text segmentation system.