|
| 基于BERT-BiGRU-CRF的电子简历命名实体识别 |
| named entity recognition;Chinese electronic resume;BERT;BiGRU;conditional random field |
| |
| DOI：10.16018/j.cnki.cn32-1650/n.202203008 |
| 中文关键词: 命名实体识别 中文电子简历 BERT BiGRU 条件随机场 |
| 英文关键词: named entity recognition Chinese electronic resume BERT BiGRU conditional random field |
| 基金项目: |
|
| 摘要点击次数: 45 |
| 全文下载次数: 52 |
| 中文摘要: |
| 针对现有的中文电子简历命名实体识别任务中，模型在训练过程中出现字向量表示单一和无法较好地处理字的多义性问题，提出了一种基于BERT的中文电子简历命名实体识别模型。该模型将电子简历的文本数据通过BERT进行字符级编码，根据上下文语境丰富字的语义信息，有效解决了一词多义的问题；将得到的字向量输入到双向门控循环单元进行特征提取，获取文本的上下文特征；利用条件随机场进行解码并生成标签序列，得到命名实体。实验结果表明，该模型能够有效提高中文电子简历命名实体识别的准确率。 |
| 英文摘要: |
| For the existing Chinese electronic resume named entity recognition task， the model presents single word vector representation and can not better handle the ambiguity of words in the training process. This paper presents a Chinese electronic resume named entity recognition model based on BERT. This model uses BERT to encode the text data of the electronic resume， enriches the semantic information of the word according to the context， and effectively solves the problem of word polysemy. The resulting word vector is input into the two-way gated loop unit for feature extraction to get the context features of the text. The conditional random field is used to decode and generate tag sequences to get named entities. The experimental results show that the model can effectively improve the accuracy of Chinese electronic resume named entity recognition. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |