Abstract: Multi-document summarization is a technology of natural languages processing，which extracts important information from multiple texts about same topic according to ratio of compression.It can execute data mining of Internet information from the global perspective.In the face of rapid growth of network resources，how to process automatic text summarization accurately and efficiently from mass data source is a main challenge in multi-document summarization.MapReduce is a distributed and parallel?computing method recommend by Google which can be deployed in cluster of any ordinary commercial computers.It can coordinate compute tasks of each computer in cluster，take full advantage of the processing power of computer cluster and analyze mass data efficiently.This paper presents an effective experimental model，which implements multi-document automatic summarization technology with MapReduce，which is a distributed and parallel framework.The results show that MapReduce can effectively improve the performance in the processing of extracting abstracts in the premise of the quality of summarization.

Key words: multi-document automatic summarization, MapReduce, parallel?computing

摘要： 多文档文摘是将同一主题下的多个文本描述的主要的信息按压缩比提炼为一个文本的自然语言处理技术，它可以从全局的角度对网络信息进行挖掘。在面对飞速增长的网络资源时，如何准确、高效地从海量数据源内进行自动文摘处理，是多文档自动文摘面临的主要难题之一。MapReduce是Google提出的一种分布式并行计算方法，它可以部署在任意一个普通商用计算机组成的集群上，能够有效地协调集群内各计算机的计算任务，充分利用计算机集群的处理能力，能够对海量数据进行有效的分析处理。提出了一个有效的实验模型，将MapReduce分布式并行框架应用在多文档自动文摘技术中。实验结果表明，MapReduce在保证文摘质量的前提下，能够有效地提高文摘抽取过程的处理性能。

关键词: 多文档自动文摘, MapReduce, 并行计算