Chinese-Vietnamese Cross-Lingual Event Causality Identification Based on Syntactic Graph Convolution

The Chinese-Vietnamese cross-lingual event causality identification aims to identify the cause and effect events from the news text describing the event information and present them in a structured form. The existing event causality extraction model faces the following two challenges: 1) The research work related to event causality extraction is mainly focused on resource-rich monolingual scenarios, and the performance of resource-scarce languages needs to be further improved; 2) Existing event causality identification methods are not good at capturing implicit causal semantic relations. Therefore, we propose a novel Chinese-Vietnamese Cross-lingual Event Causality Identification Based on Syntactic Graph Convolution. Firstly, the Chinese-Vietnamese word vectors are mapped into the same semantic space through pre-trained cross-lingual word embeddings. Then the syntactic graph convolutional neural network is used to capture the deep semantic information of the event sentence. Finally, the in-depth semantic features of event sentences in different languages are obtained by combining the cross-attention mechanism of event types. Experiment results on a self-built dataset that the proposed method outperforms the state-of-the-art models.