基于深度学习的心理疾病的情感分析模型研究.pdfVIP

长，信息通信

DOI:10.20153/j.issn.2096-9759.2025.06.018

基于深度学习的心理疾病的情感分析模型研究

赵华1.2.3,李浩1.3*，刘梦1

1河北师范大学中燃工学院，河北石家庄050024；

2河北师范大学数学学院，河北石家庄050024；

3河北省无线传感器网络数据应用技术创新中心，河北石家庄050024

ResearchonSentimentAnalysisModelsforPsychologicalDisordersBased

onDeepLearning

ZHAOHua1.2.3,IHao1.3*,LIUMengyan

1CollegeofEnergyandPowerEngineering,HebeiNormalUniversity,Shijiazhuang050024,Hebei,China;

2CollegeofMathematics,HebeiNormalUniversity,Shijiazhuang050024,Hebei,China;

3HebeiProvinceInnovationCenterforWirelessSensorNetworkDataApplication,Shijiazhuang050024,Hebei,China

摘要：该文利用深度学习模型进行情感Abstract:Thispaperemploysadeeplearningmodelforsentimentanalysisandproposesahybrid

分析，提出了一种融合BERT和深度卷modelthatintegratesBERTandDeepPyramidConvolutionalNeuralNetworks(DPCNN)to

积神经网络（DPCNN）的混合模型，用于analyzetheemotionaltendenciesintextdatarelatedtopsychologicaldisorders.Themodelutili-

分析心理疾病相关文本数据的情感倾zesdynamicwordvectorsgencratedbyBERTtocapturerichcontextualinformationandapplies

向。该模型利用BERT生成的动态词向positionalencodingtoemotionalwordstopreservetheirpositionalinformation,therebybetter

量可以学习到上下文信息并更新，对情graspingthesemanticconnectionsbetweenemotionalwords.DPCNN,bydeepeningthenet-

感词进行位置编码来保存情感词的位置worklayers,extractslong-rangedependencyrelationshipswithintextualcontextstocapturekey

信息，以便更好的把握情感词之间的语featuresfromemotionalwords.Themodelultimatelypredictstheprobabilitiesofdifferentpsy-

义联系；DPCNN通过不断加深网络，可chologicaldisordercategoriesthroughaSoftmaxlayer.Comparativeexperimentswithother

以抽取长距离的文本依赖关系，来抓取BERT-basedneuralnetworkhybridmodelsdemonstratethesuperiorperformanceoftheBERT-

情感词中的关键特征。最终模型通过DPCNNmodel,andsignificantimprovementsinaccuracy,precision,recall,andF1scores,w