A summarization system for Chinese news from multiple sources

This paper will propose a personal news secretariat that helps on-line readers absorb news information from multiple sources. Such a news secretariat eliminates the redundant information in the news, and reorganizes the news for readers. This multiple document summarization employs named entities and other signatures to cluster news stream; employs punctuation marks, linking elements, and topic chains to identify the meaningful units (MUs); employs nouns and verbs to find the similarity of MUs; and finally employs focusing and browsing models to display the summarization results. On the average, the document reduction rates are 70.77% and 42.26% for focusing and browsing summarization, respectively. The reading-time reduction rate is 30.86%, and the correct rate of question-and-answering task is 88.46% for browsing summarization.