Abstract: Aiming at the problem that the event flank is restricted by the frame structure in the news information extraction based on event frame, a method of event new flank detection is proposed, which defines two types of event new flank. The LSA clustering detection is implemented by filtering the existed flank content. The word pair feature model is used to take full use of semantic information in content-limited paragraph. This method is tested in prototype system, and experimental results show it is effective.

Key words: new flank detection, event frame, word pair feature, information extraction

摘要： 在基于事件框架的新闻信息抽取中，针对事件侧面被框架结构所限定的问题，提出一种事件新侧面探测方法，并定义事件新侧面的2种类型。通过去除已有的侧面内容，实现LSA聚类探测，同时在文本特征选取部分采用词对特征模型，以充分利用有限文本中的语法信息。在原型系统中对该方法进行测试，实验结果表明，该方法是有效的。

关键词: 新侧面探测, 事件框架, 词对特征, 信息抽取

CLC Number: 

• TP311.5