Scholarlink

Question-answering models offer much promise for improving support systems for mental health. The current work evaluates the effectiveness of RoBERTa and ELECTRA models in depression-related question answering, leveraging both the SQuAD dataset and a custom one tailored to this domain. The main metrics used in evaluation are accuracy, precision, recall, and F1-score. While both models show strong performance on the SQuAD dataset, with RoBERTa achieving 92% accuracy and ELECTRA 99%, their accuracy dropped to 45% on the custom dataset. This difference shows that there is a fundamental difficulty in adapting general-purpose models to specific applications. Our results highlight the need for domain-specific training and fine-tuning to enhance the relevance of question-answering systems in depression management contexts.

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

A Comparative Analysis of Deep Learning Models to Enhance Question Answering in Depression Management

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Preeti Tiwari,
  • Kiran Kamatham,
  • Sri Khetwat Saritha,
  • Geethika Sri Natha

摘要

Question-answering models offer much promise for improving support systems for mental health. The current work evaluates the effectiveness of RoBERTa and ELECTRA models in depression-related question answering, leveraging both the SQuAD dataset and a custom one tailored to this domain. The main metrics used in evaluation are accuracy, precision, recall, and F1-score. While both models show strong performance on the SQuAD dataset, with RoBERTa achieving 92% accuracy and ELECTRA 99%, their accuracy dropped to 45% on the custom dataset. This difference shows that there is a fundamental difficulty in adapting general-purpose models to specific applications. Our results highlight the need for domain-specific training and fine-tuning to enhance the relevance of question-answering systems in depression management contexts.