Scholarlink

<p>Mental health has gained significant attention as awareness grows about its crucial role in well-being. However, accessing effective and affordable mental health assessments and treatments remains challenging. While traditional mental health assessments often present barriers, social media has emerged as a potential goldmine of information about individuals’ psychological states. This paper presents a new depression detection method for social media posts called Deep Linguistic Analysis for Depression (DLAD). DLAD integrates the Robustly Optimized BERT Pretraining Approach (RoBERTa) and the Convolutional Neural Network (CNN) to detect depression symptoms in social media posts. RoBERTa captures nuanced contextual information and semantic relationships within the text, while the CNN model employs hierarchical feature learning to enhance the feature extraction process from RoBERTa embeddings. Several experiments are conducted to evaluate the performance of DLAD. The comparative results demonstrated the superiority of DLAD compared to alternative methods. DLAD achieved significant average improvements of 12.3%, 57.9%, 10.9%, and 34.8% in accuracy, recall, precision, and F1-score.</p>

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

Deep linguistic analysis for depression in social media using RoBERTa and CNN

  • 期刊论文

  • 发表日期：

    2025年10月3日

查看全文
  • Kawthar Salameh,
  • Tasneem Suboh,
  • Rawan ElAmayreh,
  • Bushra Alhijawi

摘要

Mental health has gained significant attention as awareness grows about its crucial role in well-being. However, accessing effective and affordable mental health assessments and treatments remains challenging. While traditional mental health assessments often present barriers, social media has emerged as a potential goldmine of information about individuals’ psychological states. This paper presents a new depression detection method for social media posts called Deep Linguistic Analysis for Depression (DLAD). DLAD integrates the Robustly Optimized BERT Pretraining Approach (RoBERTa) and the Convolutional Neural Network (CNN) to detect depression symptoms in social media posts. RoBERTa captures nuanced contextual information and semantic relationships within the text, while the CNN model employs hierarchical feature learning to enhance the feature extraction process from RoBERTa embeddings. Several experiments are conducted to evaluate the performance of DLAD. The comparative results demonstrated the superiority of DLAD compared to alternative methods. DLAD achieved significant average improvements of 12.3%, 57.9%, 10.9%, and 34.8% in accuracy, recall, precision, and F1-score.