Scholarlink

<p>Sentiment analysis is a crucial task in natural language processing that aims to extract opinions and emotions from textual data. Traditional deep learning models, such as CNNs and RNNs, have demonstrated effectiveness in sentiment classification but struggle with capturing complex emotional hierarchies, long-range dependencies, and rotational invariance in feature representations. To overcome these limitations, we propose a novel sentiment analysis framework that integrates a Bidirectional Emotional Recurrent Unit (BiERU) with a Capsule Network (CapsNet). The BiERU model efficiently captures sequential emotional dependencies, while the Capsule Network preserves hierarchical relationships between emotional components using dynamic routing, thereby improving feature robustness and classification accuracy. The proposed BiERU-CapsNet model is evaluated on the IEMOCAP dataset, demonstrating superior performance compared to existing CNN-LSTM-based architectures. Our approach achieves an accuracy of 88.5% and an F1-score of 88.6%, outperforming traditional deep learning models. Furthermore, we introduce a new dataset from AICTE’s grievance system, providing a benchmark for sentiment analysis in the education sector. The results indicate that CapsNet enhances sentiment classification by capturing complex linguistic structures, including sarcasm, negation, and mixed sentiments. This study establishes the effectiveness of integrating Capsule Networks with BiERU for sentiment analysis and opens avenues for future research in multimodal sentiment recognition across text, audio, and video domains.</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

BiERU-CapsNet: Bidirectional Emotional Recurrent Unit with Capsule Network, a Novel Approach for Textual Sentiment Analysis

  • 期刊论文

  • 发表日期：

    2026年2月6日

查看全文
  • Manoj Singh,
  • Subhash Panwar,
  • Sanju Choudhary

摘要

Sentiment analysis is a crucial task in natural language processing that aims to extract opinions and emotions from textual data. Traditional deep learning models, such as CNNs and RNNs, have demonstrated effectiveness in sentiment classification but struggle with capturing complex emotional hierarchies, long-range dependencies, and rotational invariance in feature representations. To overcome these limitations, we propose a novel sentiment analysis framework that integrates a Bidirectional Emotional Recurrent Unit (BiERU) with a Capsule Network (CapsNet). The BiERU model efficiently captures sequential emotional dependencies, while the Capsule Network preserves hierarchical relationships between emotional components using dynamic routing, thereby improving feature robustness and classification accuracy. The proposed BiERU-CapsNet model is evaluated on the IEMOCAP dataset, demonstrating superior performance compared to existing CNN-LSTM-based architectures. Our approach achieves an accuracy of 88.5% and an F1-score of 88.6%, outperforming traditional deep learning models. Furthermore, we introduce a new dataset from AICTE’s grievance system, providing a benchmark for sentiment analysis in the education sector. The results indicate that CapsNet enhances sentiment classification by capturing complex linguistic structures, including sarcasm, negation, and mixed sentiments. This study establishes the effectiveness of integrating Capsule Networks with BiERU for sentiment analysis and opens avenues for future research in multimodal sentiment recognition across text, audio, and video domains.