Scholarlink

Human communication and comprehension depend heavily on emotions, and applications like sentiment analysis, mental health evaluation, and human-computer interaction have made it more and more necessary to analyze emotional content in text data. A new dataset is introduced for the analysis of emotions during pandemic situation it is named as “PanEmo” dataset. A specialized version of BERT called EmoBERT works especially well on tasks involving emotions. In this study, we suggest an association clustering method that uses EmoBERT encoding to effectively classify text data according to emotions. To investigate the structural distribution of emotional representations, the approach combines several clustering algorithms, such as Gaussian Mixture Models (GMM), K-Means, and Hierarchical Clustering. With an Adjusted Rand Index (ARI) of 0.438, which represents the closest alignment with ground truth emotion labels, Hierarchical Clustering outperformed the others. The current research lends credence to the fact that hierarchical methodologies are sufficient in modelling complex emotional sequences. In general, the paper makes contributions to sentiment analysis and natural language processing, and it may have implications on both emotion-aware chatbots and sentiment-based recommendation systems.

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

PanEmo: Decoding Emotions During Pandemic from Social Media Tweets Using Unsupervised Clustering Methods

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Soubraylu Sivakumar,
  • Abhivesh Shukla,
  • Prantik Ghosh,
  • Selvanayaki Kolandapalayam Shanmugam,
  • Ratnavel Rajalakshmi,
  • P. Selvaraju

摘要

Human communication and comprehension depend heavily on emotions, and applications like sentiment analysis, mental health evaluation, and human-computer interaction have made it more and more necessary to analyze emotional content in text data. A new dataset is introduced for the analysis of emotions during pandemic situation it is named as “PanEmo” dataset. A specialized version of BERT called EmoBERT works especially well on tasks involving emotions. In this study, we suggest an association clustering method that uses EmoBERT encoding to effectively classify text data according to emotions. To investigate the structural distribution of emotional representations, the approach combines several clustering algorithms, such as Gaussian Mixture Models (GMM), K-Means, and Hierarchical Clustering. With an Adjusted Rand Index (ARI) of 0.438, which represents the closest alignment with ground truth emotion labels, Hierarchical Clustering outperformed the others. The current research lends credence to the fact that hierarchical methodologies are sufficient in modelling complex emotional sequences. In general, the paper makes contributions to sentiment analysis and natural language processing, and it may have implications on both emotion-aware chatbots and sentiment-based recommendation systems.