摘要
文本分类技术能够帮助心理咨询对话系统自动判别用户的心理状态,以便在聊天过程中正确对用户进行心理治疗及心理健康干预,在心理学领域中具有良好的应用前景。本文在近年提出的Emotional First Aid Dataset心理咨询语料库上依次构建了烦恼类型、心理疾病、伤害身体倾向三个文本多分类任务,提出了该语料库的数据预处理方案,同时研究了BERT、Ro BERTa等6个深度学习语言模型在这些多分类任务上的性能,并以这些模型作为基学习器构建了集成模型。实验结果表明,XLNet、RoBERTa、ERNIE模型在多个任务上的表现较为突出,同时集成学习能显著地提高分类模型的预测准确率,整体取得了良好的效果。
Text classification technology can help the psychological counseling dialogue system to automatically identify the user's psychological state,so as to correctly provide psychological treatment and mental health intervention to the user during the chat process,and has a good application prospect in the field of psychology.Based on the Emotional First Aid Dataset proposed in recent years,this paper sequentially constructs three text multi-classification tasks of types of annoyance,mental illness,and tendency to harm the body,and proposes a data preprocessing scheme for the corpus.At the same time,this paper studies the performance of 6 deep learning for language modeling such as BERT and RoBERTa on these multi-classification tasks,and builds an integrated model using these models as the base learner.The experimental results show that the XLNet,RoBERTa,and ERNIE models perform well on multiple tasks.In addition,ensemble learning can significantly improve the prediction accuracy of the classification model,and overall good results have been achieved.
作者
林子洛
LIN Ziluo(School of Mathematical Sciences,South China Normal University,Guangzhou Guangdong 510631)
出处
《软件》
2023年第7期112-118,共7页
Software
关键词
深度学习语言模型
心理学领域文本分类
集成学习
deep learning for language modeling
text classification in the field of psychology
ensemble learning
