提出一种基于浅层分析的多文档文摘方法,该方法分析了单文档的结构信息,多文档的统计信息,并利用改进的MMR方法动态地计算文摘候选句子的加入对文摘的贡献,去除冗余信息,最终按照一定时间顺序输出多文档文摘.对生成的英文文摘进行基于n-gram方法的自动评测,结果表明该方法具有较好的信息覆盖率,具有一定的实用价值.
An approach of multi-document summarization based-on shallow analysis is presented. It analyzes the structure information of the single document and statistical information of multi-document, and dynamically computes the contribution of a sentence to the summary and reduces the redundant information with improved MMR method. Then the summary outputs with a timeline. The automatic evaluation of the English summaries with n-gram score shows the system's effectiveness and feasibility.