摘要
针对深度学习模型特征提取不全面、静态词向量无法表示多义词等问题,提出基于多特征融合网络的电力设备缺陷文本分类模型。利用ALBERT模型动态调整词在具体上下文语境中的向量表示,获取电力文本的动态特征表示;利用多特征融合网络MCNN-TCN模块充分提取文本中的局部特征和上下文语义联系,注意力机制赋予模型聚焦于关键特征的能力。在真实电力设备缺陷文本数据集进行实验,结果表明基于多特征融合网络分类模型具有更好的分类效果,模型F1分数达到了97.28%。
To address the problems that the feature extraction of deep learning model is not comprehensive and the static word vector can not solve polysemy,a power equipment defect text classification model based on multi feature fusion network is proposed.The ALBERT is used to dynamically adjust the vector representation of words in the specific context to obtain the dynamic feature representation of power text.The multi-feature fusion network MCNN-TCN fully extracts the local features and context semantic relations,and the attention mechanism gives the model the ability to focus on key features.Experiments on real power equipment defect text dataset show that the multi-feature fusion network classification model has better effect,and the F1 score reaches 97.28%.
作者
赵瑞锋
李波
卢建刚
李世明
曾坚永
郑文杰
ZHAO Ruifeng;LI Bo;LU Jiangang;LI Shiming;ZENG Jianyong;ZHENG Wenjie(Electric Power Dispatching and Control Center of Guangdong Power Grid Co.,Ltd.,Guangzhou 510600,China)
出处
《微型电脑应用》
2023年第7期81-84,共4页
Microcomputer Applications
