Joint learning for disaster event extraction
Chao Li, Yan Jiang · 6th International Workshop on Advanced Algorithms and Control Engineering (IWAACE 2022) · 2022
Natural disasters such as forest fires, earthquakes, and droughts will have a huge impact on people's lives and the social economy, which makes people and governments pay great attention to these disasters. Therefore, the automatic extraction of disaster events and relevant information about events has high research value. The event extraction task based on Pipeline will have cascading errors and the use of natural language processing tools in the event extraction task will cause the problem of error accumulation. This paper proposes a joint event extraction method(BERT-LSTM-GRAPH, BLG) for disasters. This method converts the task of event extraction into a sequence labeling task and uses a graph convolutional neural network to fuse character information and word information to complete the joint extraction of triggers and arguments. Experiments show that the BLG model can improve the extraction performance of triggers and arguments in the field of disasters.