基于概念图谱与BiGRU-Att模型的突发事件演化关系抽取

针对现有突发事件演化关系抽取方法仅利用了句子本身的信息而忽略了背景知识的问题,引入概念图谱进行特征拓展,以获得更多的语义信息来改善演化关系抽取效果。首先根据概念图谱构建突发事件语义网络,通过特征迭代选择算法获得演化因子的概念特征,然后联合概念特征与突发事件文本作为双向门控循环单元(BiGRU)模型的输入进行特征学习,并利用注意力(Attention)机制对输出的特征信息序列实施加权变换,最后将学习到的特征序列输入到分类器进行演化关系分类。实验结果表明,所提出的基于概念图谱与BiGRU-Att模型的方法相比于现有方法,在准确率、召回率和F_1值等评价指标上均有提升。

Evolution relationship extraction of emergency based on concept graph and BiGRU-Att model

To solve the problem that the existing methods for evolution relationship extraction of emergency only use the context of the sentences but ignore the background knowledge, this study introduces concept graph to expand the semantic features, so as to obtain more semantic information and improve the performance of evolution relationship classification. Firstly, an emergency semantic network is built based on concept graph, and concept features of the evolution factors are chosen by an iterative scheme. Secondly, the concept features combined with the emergency text are input to the bidirectional gated recurrent unit (BiGRU) model for feature learning, and the attention mechanism is used to transform the weight of the output feature sequence. Finally, the feature sequence is input to a classifier for evolution relationship classification. The experimental results show that, to judge by such evaluation indexes as precision, recall and F₁-score, the proposed method based on concept graph and BiGRU-Att model is superior to the existing ones.