基于BERT-BiGRU-CRF的突发事件元素抽取模型研究

Research on BERT-BiGRU-CRF-based emergency extraction model

针对中文突发事件领域的事件元素抽取准确率低的问题,文章提出一种BERT-BiGRU-CRF突发事件元素抽取模型,旨在将事件元素抽取任务转变为序列标注任务以进行处理。其中,首先通过BERT预训练实现突发事件文本向量化;然后通过BiGRU进一步实现语义编码;最后采用CRF研判输出最大概率标注序列。对比结果证明,BERT-BiGRU-CRF模型具有更高的抽取准确率,可以在突发事件元素抽取任务中获得更好的抽取效果。

In response to the low accuracy of event element extraction in the field of Chinese emergency events,this article proposes a BERT-BiGRU-CRF emergency event element extraction model,aiming to transform the task of event element extraction into a sequence annotation task for processing.Firstly,the text vectorization of sudden events is achieved through BERT pre training;Then further implement semantic encoding through BiGRU.Finally,CRF is used to determine the maximum probability annotated sequence output.The comparative results demonstrate that the BERT-BiGRU-CRF model has higher extraction accuracy and can achieve better extraction results in emergency event element extraction tasks.