摘要
基于抑郁情绪人群具有生理行为和语言行为的双层次特征维度特点,结合数学模型与机器学习的方法对抑郁情绪的识别展开实验和研究。提出利用熵值层次分析法判断有抑郁情绪人群的日常行为特征,如情绪低落周期、极端行为等,并结合二次特征词库加强的方式,得到更为合理特征词向量并输入到贝叶斯算法,达到对抑郁情绪进行文本识别的目的。实验结果表明,从这两个维度进行抑郁情绪症状的识别方法能有效地提高识别效果。
Based on the two-level feature dimension characteristics of physiological behavior and language behavior of depressed people,this paper combines mathematical models and machine learning methods to carry out experiments and research on the recognition of depressed emotions.It proposes to use the entropy analytic hierarchy process to judge the daily behavior characteristics of people with depression,such as depression cycle,extreme behavior,etc.;and combines with the secondary feature vocabulary to strengthen the TF-IDF algorithm,the obtained feature word vector is input into the Bayesian algorithm to achieve the purpose of text recognition of depressive emotions.The experimental results indicate that the recognition of depression and emotional symptoms from two dimensions can effectively improve the recognition effect.
作者
宋锐彪
SONG Ruibiao(Yunnan Normal University,Kunming Yunnan 650500,China)
出处
《通信技术》
2022年第2期187-192,共6页
Communications Technology
基金
云南师范大学大学生科研训练基金项目。
关键词
抑郁情绪
熵值层次分析法
词频-逆文档频率算法
贝叶斯模型
depressed emotion
entropy analytic hierarchy process
TF-IDF(Term Frequency-Inverse Document Frequency)algorithm
Bayesian model
