【摘要】 为了提高命名实体识别模型的系统实用性,有效利用互联网中海量未经标注的数据,提出了一种基于多神经网络协同训练的命名实体识别模型。该模型融合了循环神经网络和协同训练的优势,首先利用少量的有标记数据训练3种不同的神经网络获得初始识别模型,然后在大量无标注数据上对3种神经网络模型进行协同训练以优化模型。实验结果表明,本文模型能够有效地训练大量的无标记数据,与传统的协同训练和单一神经网络识别模型相比,模型的整体性能得到了显著提升。更多还原

【Abstract】 In order to improve the system practicability of the named entity recognition model and effectively utilize the massive unlabeled data in the Internet,this paper proposes a named entity recognition model based on tri-training of multi-neural network.The model combines the advantages of recurrent neural network and tri-training. Firstly,three different neural networks are trained with a small amount of labeled data to obtain the initial recognition model,then the tri-training of three neural network named entity recognition models are performed on a large number of unlabeled data to optimize the model. The experimental results show that the model can effectively train a large amount of unlabeled data,and the overall performance of the model is significantly improved compared with the traditional tri-training and single neural network recognition model.更多还原