Generating the Theme Overview Based on Clue Chain from Online News

The online news media has been an important resource that people digest information. However, people may be overwhelmed by the huge number of online news, making it hard to get the theme development overview of a news topic. In this paper a new method is proposed to generate the theme overview based on the clue chains from a large number of relevant online news. Given a group of relevant articles for a news topic, the proposed method constructs original clue chains after identified theme clue sentences by the conditional random field model and then filters original chains to get theme clue chains by taking time and semantic similarity between clue chains as conditions. Each clue can be represented by its time and strength and the clues with high strength are selected as theme foci. For each theme focus, the relevant theme clue sentences corresponding to the clue chain are selected to form a summary. All summaries are then ordered chronologically to form the theme overview of the news. Experiment result shows the effectiveness of our approach and the theme overview of news obtained can explicitly depict the news development trend.