摘要
为提升医疗服务质量,提出一种基于深度语义分析与满意度预测的医患沟通文本评价方法。该方法采用多任务建模框架,融合Transformer架构,利用预训练语言模型提取文本语义特征,并结合情感分析任务引导模型学习文本中的主观情绪信息,最终通过回归任务实现满意度预测。为进一步提升模型性能的稳定性与预测精度,引入量子粒子群算法对关键超参数进行全局优化。实验结果表明,在情感分类任务中,模型AUC达到0.921,情感分类准确率为85.27%;在满意度预测任务中,MSE和MAE分别为0.559与0.436,预测准确率达81.40%,多项性能指标显著优于基线模型。研究表明,所提方法能有效挖掘医患沟通文本中的深层语义关联与情感倾向,在医患沟通评价与医疗服务优化方面具有较好的应用潜力。
To improve the quality of medical services,a text evaluation method for doctor-patient communication based on deep semantic analysis and satisfaction prediction is proposed.This method adopts a multi-task modeling framework,integrates the Transformer architecture,uses pre-trained language models to extract semantic features of the text,and combines sentiment analysis tasks to guide the model to learn subjective emotional information in the text.Ultimately,satisfaction prediction is achieved through regression tasks.To further enhance the stability of model performance and prediction accuracy,the study introduces the quantum particle swarm optimization algorithm to globally optimize key hyperparameters.The experimental results show that in the emotion classification task,the AUC of the model reaches 0.921 and the accuracy rate is 85.27%.In the satisfaction prediction task,the MSE and MAE were 0.559 and 0.436 respectively,with a prediction accuracy rate of 81.40%.Many performance indicators were significantly better than those of the baseline model.Research shows that the proposed method can effectively mine the deep semantic associations and emotional tendencies in doctor-patient communication texts,and has good application potential in the evaluation of doctor-patient communication and the optimization of medical services.
作者
任昱君
黄先涛
宋晶晶
金宣伯
贾丽群
REN Yujun;HUANG Xiantao;SONG Jingjing;JIN Xuanbo;JIA Liqun(Department of Legal Affairs,The First Affiliated Hospital of Hebei North University,Zhangjiakou 075000,China)
出处
《电子设计工程》
2026年第5期12-16,共5页
Electronic Design Engineering
基金
北京市自然科学基金(2017D01C120)。
