融合深度学习与规则的民族工艺品领域命名实体识别

Named Entity Recognition in Ethnic Handicraft Field with the Deep Learning and Rules

提出了一种融合深度学习与规则的民族工艺品领域实体识别方法.首先通过BERT预训练语言模型获得语义向量;然后将其输入到BiLSTM-CRF序列标注模型中训练并预测初步结果;最后根据领域特点提出相应的规则对错误预测结果校正.实验结果表明,在自建的民族工艺品数据集上获得的准确率、召回率和F1值分别为95.43%、90.88%和93.10%,可以有效地提取民族文本中命名实体信息.

A method of combining deep learning and rules is proposed for Named Entity Recognition in ethnic crafts field.The first will obtain the semantic vector through the BERT(Bidirectional Encoder Representations from Transformers)pre-training language model;Then input it into the BiLSTM-CRF sequence labeling model to train and predict preliminary results;Finally,according to the characteristics of the field,the corresponding rules are proposed to correct the error prediction results.The experimental results show that the accuracy,recall and F1 values obtained from the self-built ethnic craft data set are 95.43%,90.88% and 93.10% respectively,this model can extract name entity of ethnic text efficiently.