Scholarlink

The main aim of our research is to automatically discretize the symptoms of depression expressed on social networks, in accordance with the 21 items of the BDI-II (Beck Depression Inventory) questionnaire. This discretization aims to assign each detected symptom a discrete level of severity (from 0 to 3), consistent with the BDI-II scale, based on text messages posted online. Our approach relies on two complementary steps: the extraction of linguistic expressions associated with depressive symptoms, followed by their discretization using statistical methods such as KMeans clustering and quantile segmentation. This combination transforms raw textual signals into interpretable numerical representations of the degree of psychological distress. By leveraging natural and spontaneous data from users’ everyday lives, our system goes beyond the limitations of traditional mental health diagnostic methods. It enables automatic, fine-grained, and continuous assessment of symptom severity, paving the way for the early detection of individuals at risk. Ultimately, this system could serve as a valuable decision-support tool for mental health professionals.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automatic Discretization of Depression Symptoms from Social Media Using KMeans and Quantiles

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hathemi Mohamed,
  • Manel Ben Amira,
  • Nabil Khoufi,
  • Chafik Aloulou

摘要

The main aim of our research is to automatically discretize the symptoms of depression expressed on social networks, in accordance with the 21 items of the BDI-II (Beck Depression Inventory) questionnaire. This discretization aims to assign each detected symptom a discrete level of severity (from 0 to 3), consistent with the BDI-II scale, based on text messages posted online. Our approach relies on two complementary steps: the extraction of linguistic expressions associated with depressive symptoms, followed by their discretization using statistical methods such as KMeans clustering and quantile segmentation. This combination transforms raw textual signals into interpretable numerical representations of the degree of psychological distress. By leveraging natural and spontaneous data from users’ everyday lives, our system goes beyond the limitations of traditional mental health diagnostic methods. It enables automatic, fine-grained, and continuous assessment of symptom severity, paving the way for the early detection of individuals at risk. Ultimately, this system could serve as a valuable decision-support tool for mental health professionals.