摘要
学生心理健康问题是学校与家庭关注的焦点,情感分析模型也随之被广泛应用于对学生心理健康的干预中。现有情感模型在复杂情感识别与多模态数据融合方面存在局限,故提出一种基于Transformer网络的多模态情感分析模型,采用LSTM模块提取语音与视觉特征,结合BERT提取文本特征,利用Transformer网络实现多模态特征深度融合。实验在CMU-MOSI数据集上进行,结果表明,模型在MAE、Corr与Acc-7等指标上显著优于传统模型,展现出较高的准确性与鲁棒性。基于Transformer网络的情感分析模型可解释性及在学生心理健康管理中的应用具备较大潜力,可以为学生心理健康干预提供有效的技术支持。
Students’mental health problems are becoming more and more serious.Emotion analysis model is gradually applied to mental health intervention as technical means.However,the existing models are limited in complex emotion recognition and multi-modal data fusion,and it is difficult to meet the actual needs.Therefore,the study proposes a multi-modal sentiment analysis model combined with BERT to extract text features,which uses LSTM module to extract speech and visual features,and uses Transformer network to achieve deep integration of multi-modal features.Experiments were carried out on the CMU-MOSI dataset.The results showed that the model was significantly better than the traditional method in MAE,Corr and Acc-7,showing higher accuracy and robustness.The study also discusses the interpretability of the model and its potential application in students’mental health management to provide effective technical support for mental health intervention.
作者
罗斌玲
陈光会
王文玉
Luo Binling;Chen Guanghui;Wang Wenyu(Guangxi College of Water Resources and Electric Power,Nanning 530023,China)
出处
《黑龙江科学》
2025年第5期106-108,共3页
Heilongjiang Science
基金
2024年度广西高校中青年教师科研基础能力提升项目“情感分析模型在学生心理健康干预中的应用研究”(2024KY1154)。
