现代电子技术2025，Vol.48Issue(13)：123-132,10.DOI:10.16652/j.issn.1004-373x.2025.13.018
融合双通道特征信息的医疗短文本分类模型
Medical short text classification model with fusion of dual channel feature information
摘要
Abstract
In view of the sparse features,semantic ambiguities and insufficient extraction of short text features in the medical short texts,a medical short text classification model EBDF(ERNIE-BiLSTM-DPECNN-FGM)fusing dual-channel features is proposed.The pre-trained model is used to obtain dynamic word vectors,which made the model contain richer semantic information.Then the BiLSTM is used to obtain global text feature information and the DPECNN is used to obtain deep local text feature information.The FGM adversarial training algorithm is used to disturbance the data to improve the robustness and generalization ability of the model.Finally,the feature information of the two channels is fused to obtain the final text representation.The EBDF model was compared with the model with the better effect on the short text data sets of three medical fields and two general fields.It can be seen that its accuracy is improved by about 0.57%～6.16%,and its F1 value is improved by about 0.65%～5.80%.关键词
医疗文本挖掘/短文本分类/特征融合/BiLSTM/DPECNN/双通道Key words
medical text mining/short text classification/feature fusion/BiLSTM/DPECNN/two-channel分类
信息技术与安全科学引用本文复制引用
李晨,刘纳,郑国风,杨杰,道路..融合双通道特征信息的医疗短文本分类模型[J].现代电子技术,2025,48(13):123-132,10.基金项目
宁夏自然科学基金项目(2021AAC03224) （2021AAC03224）
北方民族大学校级科研项目(2024XYZJK01) （2024XYZJK01）
北方民族大学研究生创新项目(YCX23167) （YCX23167）
