IPS - Modified DSC-TextRank method of automatic text summarization - Atos

Abstract

Automatic text summarization is a promising area of natural language processing. Today, information is flooding on the Internet. Thus, the summarization of key ideas from vast materials (e.g. news) becomes quite necessary. So far, there is still not many mature applications in Chinese text summarization. So, this project is aimed to construct an application of Chinese text summarization. The difficulty in summarizing Chinese texts is the texts have to be segmented into separate words first. Since Chinese segmentation is a very professional topic, this project will emphasize on the processing algorithm after segmentation and leave the segmentation done by external tools. The core algorithm of this project is a new algorithm modified on the base of DSC (Massimiliano Pavan and Marcello Pelillo, 2007) and TextRank (Rada Mihalcea and Paul Tarau, 2004) methods. To present the effect of the proposed summarization algorithm, I set up a news website which collects Chinese news from the Internet and then display the automatic summaries. Also, the comments are processed by the proposed algorithm to show a summarized attitude. This kind of application can help the public understand the main ideas out of exploded information today in a faster and better way.