【作者】 张江英； 郝矿荣； 王直杰； 唐雪嵩； 刘肖燕； 任立红；

【Author】 Jiangying Zhang;Kuangrong Hao;Zhijie Wang;Xue-song Tang;Xiaoyan Liu;Lihong Ren;College of Information Science and Technology,Engineering Research Center of Digitized Textile &Apparel Technology,Ministry of Education,Donghua University;

【机构】 东华大学信息科学与技术学院数字化纺织服装技术教育部工程研究中心；

【摘要】 为了提高人们对危险环境变化的感知,从新闻报道中自动抽取紧急事件具有重要意义。深度神经网络虽广泛应用于事件抽取并取得了显著成果,但它们大多是基于数据集ACE2005进行的。该数据集不包含紧急事件,也不适合作为输入的实际应用。因此,本文构造了一种网格结构的长短时记忆网络(LatticeLSTM)中文紧急事件抽取模型。该模型利用预训练模型进行字符向量嵌入,并使用条件随机场(CRF)捕获触发词和事件元素间的相互作用。在CEC语料库上实验证明了该模型的有效性,且其整体性能优于其他最新方法。更多还原

【Abstract】 To improve people’s perception of changes in the dangerous environment, it is of great significance to automatically extract emergency events from news. Deep neural networks have been widely used in event extraction tasks and have achieved remarkable results, but most of them are based on the ACE2005 dataset, which does not contain emergency events and it is not suitable for practical applications as input. Therefore, we propose an Lattice LSTM model to extract Chinese emergency events, we apply a pre-trained model for character vector embedding and use a conditional random field(CRF) to capture the interaction between trigger words and event arguments. The effectiveness of the presented model is proved through experiments, and it shows that the overall performance of our model is better than other state-of-the-art methods.更多还原