基于事件时序关系的自动摘要抽取

由于文本中事件之间的时序关系可以帮助人们更好地理解文本内容,故针对新闻报道类文本,将事件作为其基本语义单元,并根据时序关系建立事件有向网络文本表示模型;利用PageRank算法结合主题相关度对时序网络进行节点重要度计算及调整;最后,按照重要度以及事件发生的顺序进行排序,并按照一定的压缩比提取摘要句,删除冗余的句子,将事件对应的原语句作为摘要。实验结果表明,基于事件时序关系的自动摘要方法效果较好。

Automatic Summary Extraction Based on Event Temporal Relationship

Since the temporal relationship between events in the text can help people better understand the content of the text, for news report texts, events are used as the basic semantic unit, and a directed network text representation model of events is established according to the temporal relationship. PageRank algorithm combined with topic relevance is used to calculate and adjust node importance of time series network. Finally, the sequence is sorted according to the importance and the sequence of events, and the abstract sentence is extracted according to a certain compression ratio, and the redundant sentence is deleted, and the original sentence corresponding to the event is taken as the summary. The experimental results show that the automatic summarization method based on event sequence relation is effective.