基于事件要素的自动文摘抽取

Automatic Extraction Summarization Based on Event Elements

对传统自动文摘技术中容易产生的信息冗余和内容覆盖不全面问题,而目前主流的技术主要是面向词语的自动文摘。论文针对事件知识粒度下的事件要素在该问题上的有效性进行研究。首先通过标注好的CEC语料库来获取事件要素,然后构建事件要素无向图,其次再对无向图节点和无向边进行权值计算,最后得到简练的文摘句,进而按照原文本顺序输出文摘。实验主要在CEC语料库上进行,较其它方法而言,召回率和准确率取得了较好的效果,平均F值可达0.62,能更好地概括文本内容。

When adopting traditional automatic summarization,it emerges information redundancy and incomplete content covering,but currently the mainstream automatic summarization turns towards to extracting words.In this paper the effectiveness of this issue about event elements on the size of event is studied.Firstly the event elements through the tagged CEC corpus are obtained;Then an event element undirected graph is built,nodes＇and edges＇weights of the undirected graph are calculated;Finally the concise summary sentences are gotten and the text summarization in accordance with the original text sequence is outputted.Experiments are conducted on CEC corpus,recall and precision have got better results to many other methods and the average value F of this method can be raised to 0.62,which can better generalize the text content.