Chinese financial event element extraction based on self-attention mechanism
To solve the problem that multiple pronouns refer to the same element in the Chinese financial event element extraction task,an event element extraction model based on self-attention mechanism was proposed.The model integrated the domain knowledge of financial events and event type knowledge in the preprocessing stage so that the pre-training model could obtain more reliable event element representations according to the event type information.Then,the multi-head attention mechanism was used to mine the referential meanings of pronouns for different elements in the news context to achieve referential resolution between overlapping elements.Finally,the bidirectional long short-term memory network and the conditional random field were used to mine the contextual feature representation of the long text of news to realize the extraction of event elements.A Chinese financial event corpus was constructed,and the effectiveness of the model was verified through comparative experiments with mainstream models.
deep learningfinancial event extractioncorpus constructionevent element extraction
万方数据
维普
