基于字词联合表示的中文事件检测方法

Chinese Event Detection with Joint Representation of Characters and Words

事件检测作为事件抽取的一个子任务,是当前信息抽取的研究热点之一。它在构建知识图谱、问答系统的意图识别和阅读理解等应用中有着重要的作用。与英文字母不同,中文中的字在很多场合作为单字词具有特定的语义信息,且中文词语内部也存在特定的结构形式。根据中文的这一特点,文中提出了一种基于字词联合表示的图卷积模型JRCW-GCN(Joint Representation of Characters and Words by Graph Convolution Neural Network),用于中文事件检测。JRCW-GCN首先通过最新的BERT预训练语言模型以及Transformer模型分别编码字和词的语义信息,然后利用词和字之间的关系构建对应的边,最后使用图卷积模型同时融合字词级别的语义信息进行事件句中触发词的检测。在ACE2005中文语料库上的实验结果表明,JRCW-GCN的性能明显优于目前性能最好的基准模型。

As a sub-task of event extraction,event detection is a research hotspot in information extraction.It plays an important role in many NLP applications,such as knowledge graph,question answering and reading comprehension.Different with English character,a Chinese character can be regarded as single-character word and has its certain meaning.Moreover,there are specific structures in Chinese words.Therefore,this paper proposes a Chinese event detection model based on graph convolution neural network,called JRCW-GCN(Joint Representation of Characters and Words by Graph Convolution Neural Network),which integrates Chinese character and word representation.It uses the latest BERT and Transformer to encode the semantic information of words and characters respectively,and then uses the relationship between words and characters to construct the corresponding edges.Finally,it uses the graph convolution model to detect Chinese events by integrating the semantic information in Chinese character and word level.The experimental results on the ACE2005 Chinese corpus show that the performance of our model outperforms the state-of-the-art models.