摘要
医疗疾病文本的准确分类对医疗信息化的发展具有重要的推进作用,本研究提出一种基于双通道学习的神经网络模型研究疾病文本分类方法。该模型分别使用卷积神经网络和双向长短期记忆网络对患者输入的疾病症状文本进行局部特征以及时序特征学习。此外,在双向长短期记忆网络上引入自注意力机制区分特征对类别预测的贡献值,增强模型的学习能力和可解释性。为使两个通道提取到的特征能够共同决定分类结果,该模型将两种特征进行拼接融合,最后利用softmax分类器得到最终的分类结果。实验结果表明,在疾病文本分类的性能方面,该模型相比其他分类模型具有较高的精确率、召回率和F1值,分别可达90.61%、90.48%和90.51%。
The accurate text classification for diseases plays an important role in promoting the development of medical informatization.A method for studying disease text classification based on a two-channel neural network model is proposed.In the model,convolutional neural network and bidirectional long short-term memory network are used to learn the local features and temporal features of disease symptom text input by patients.In addition,self-attention mechanism is introduced to bidirectional long short-term memory network for distinguishing the contribution value of the feature to classification prediction,which enhances the learning ability and interpretability of the model.The model combining two kinds of features for enabling the features extracted from the two channels to jointly determine classification results,and finally softmax classifier is used to obtain the final classification results.Experimental results show that the accuracy,recall rate and F1 value of the proposed model are 90.61%,90.48%and 90.51%,respectively,which were higher than those of other classification models.
作者
袁野
廖薇
YUAN Ye;LIAO Wei(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)
出处
《中国医学物理学杂志》
CSCD
2021年第5期655-660,共6页
Chinese Journal of Medical Physics
基金
国家自然科学基金(62001282)
上海高校青年东方学者岗位计划资助项目(QD2017043)。
关键词
疾病文本分类
自注意力
卷积神经网络
双向长短期记忆网络
disease text classification
self-attention
convolutional neural network
bidirectional long short-term memory network
