基于XLBIC的石油开采数据命名实体识别研究
Research on Named Entity Recognition of Petroleum Exploitation Data Based on XLBIC
任伟建 1计妍 2康朝海1
作者信息
- 1. 东北石油大学电气信息工程学院,黑龙江 大庆 163318;黑龙江省网络化与智能控制重点实验室,黑龙江 大庆 163318
- 2. 东北石油大学电气信息工程学院,黑龙江 大庆 163318
- 折叠
摘要
在石油领域命名实体识别的任务中,提出了基于XLBIC(XLNet-BiGRU-IDCNN-CRF)的命名实体识别模型.首先采用XLNet预训练模型获取丰富且准确的词向量信息,将获取的词向量信息送入BiGRU和IDCNN网络中进行特征提取.针对膨胀卷积网络(IDCNN)获取特征维度不高,模型计算速度较慢的问题,提出在IDCNN网络中引入门控机制,实现信息的多通道传输和流量控制,提高模型的计算速度.实验表明XLBIC命名实体识别模型在自建石油开采数据集上性能相比其它模型有提高,准确率在90%以上.
Abstract
In the task of named entity recognition in the petroleum field,a named entity recognition model based on XLBIC(XLNET-BigRU-IDCNN-CRF)is proposed.Firstly,the XLNet pretraining model was used to obtain rich and accurate word vector information,and then the obtained word vector information was sent to BiGRU and IDCNN networks for feature extraction.Aiming at the problem of low feature acquisition dimension and slow model calculation speed of the dilatative convolutional network(IDCNN),a gating mechanism was introduced in the IDCNN network to realize multi-channel information transmission and flow control and improve the model calculation speed.Experimental results show that the XLBIC named entity recognition model has better performance than other models in the self-built oil production data set,and the accuracy is more than 90%.
关键词
命名实体识别/膨胀卷积网络/门控机制Key words
Named entity recognition/IDCNN/Gating mechanisms引用本文复制引用
基金项目
国家自然科学基金资助项目(61933007)
国家自然科学基金资助项目(61873058)
黑龙江省自然科学基金(F2018004)
黑龙江省自然科学基金(F2018005)
出版年
2024
NETL
NSTL
万方数据