基于自注意力深度学习的微博实体识别研究

Research on Microblog Entity Recognition Based on Self-Attention Deep Learning

命名实体识别是自然语言处理的重要基础,随着神经网络的快速发展,深度学习的各种方法被应用于文本处理的各个方向。引入自注意力机制,结合深度学习方法,提出一种基于自注意力的双向长短期记忆条件随机场(SelfAtt-BiLSTM-CRF)方法来识别微博中的实体,利用自注意力机制,获取词与词之间的依赖关系,进一步提高模型的识别能力。实验表明,所提出的方法取得了较好的识别效果。

Named Entity Recognition(NER) is an important basis for natural language processing. With the rapid development of neural networks, various methods of deep learning have been pervasively applied to text processing.By introducing the self-attention mechanism, as well as combined with deep learning method, a self-attention-based bidirectional long-term and short-term memory conditional random field(SelfAtt-BiLSTM-CRF) method has been proposed to identify entities in microblogs. By utilizing the self-attention mechanism to obtain the dependency between words, this method helps to further improve the recognition ability of the model. Experiments show that the method proposed in this paper has achieved a satisfying recognition effect.