中文事件抽取通常使用循环神经网络(recurrent neural network, RNN)来进行事件和事件要素的抽取, 但 RNN 在处理长度较长的词语时容易丢失重要信息, 为此提出一种组合卷积神经网络(convolutional neural network, CNN)与双向长短期记忆(bidirectional long short-term memory, Bi-LSTM)网络的中文事件抽取模型 CNN-Bi-LSTM-CRF, 其中 CRF (conditional random field) 为条件随机场. 采用基于注意力机制和语义特征生成的字词联合向量, 使用 CNN 和 Bi-LSTM 模型对字词联合向量进行处理, 以获取其隐含表示, 最后通过 CRF 得出预测结果. 实验结果表明, 所提出的方法与其他现有的中文事件抽取方法相比, 准确率有明显提升.

The recurrent neural network is widely used in the event extraction of Chinese text to extract events and event elements, but it usually loses essential information when processing long words. In this study, the convolutional neural network (CNN) and the bidirectional long short-term memory (Bi-LSTM) network were combined to develop a novel event extraction model known as CNN-Bi-LSTM-conditional random field (CRF). A joint vector of characters and words was adopted based on the attention mechanism and semantic features, and the CNN and Bi-LSTM models were used to process the vector to obtain its implicit representation. Finally, the CRF was used to obtain the prediction results. The experimental results show that the proposed method is more accurate than other existing event extraction methods in extracting Chinese text.