A Chinese Event Relation Extraction Model Based on BERT

Relation extraction and event extraction are important subtasks of information extraction. To identify the relations and events in Chinese text accurately can help to improve the performance of tasks such as graph construction and risk conduction. Different from the traditional methods, this paper proposes a joint model to extract entities and events in the text, and gives the concept of event relation, to discover the potential relations between the arguments of events and the relations between two or more events. We conduct experiments on a financial dataset, the results show that the new model is 4%-6% higher than the existing event extraction model in F1 score, and the proposed event relation is also meaningful and practical.