Scholarlink

<p>This work introduces a novel sentiment classification framework designed to address vagueness, uncertainty, and noise in textual data. The proposed methodology integrates Entropy-Based Filtering with Type-2 Fuzzy Sets (T2FS) to enhance uncertainty modeling. Also, it utilizes a Gaussian Mixture Model (GMM) in conjunction with the Bayesian Information Criterion (BIC) for optimal threshold determination. Sentiment polarity is computed using SentiWordNet (SWN), while entropy-based filtering is employed to select lexically significant words contributing to sentiment evaluation. The framework is validated across six heterogeneous datasets, demonstrating consistent&#xa0;improvements in F1-score, recall, precision, and accuracy, particularly under conditions of linguistic ambiguity. These results substantiate the effectiveness and domain independence of the proposed approach for sentiment analysis and rating prediction. Furthermore, the framework specifically addresses the challenges posed by the vagueness and uncertainty present in natural language, offering a robust solution for handling ambiguous textual data in real-world applications. </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

Entropy-driven sentiment rating generation for online textual reviews with type-2 fuzzy sets

  • 期刊论文

  • 发表日期：

    2025年7月19日

查看全文
  • Divya Arora,
  • Devendra K. Tayal,
  • Sumit K Yadav

摘要

This work introduces a novel sentiment classification framework designed to address vagueness, uncertainty, and noise in textual data. The proposed methodology integrates Entropy-Based Filtering with Type-2 Fuzzy Sets (T2FS) to enhance uncertainty modeling. Also, it utilizes a Gaussian Mixture Model (GMM) in conjunction with the Bayesian Information Criterion (BIC) for optimal threshold determination. Sentiment polarity is computed using SentiWordNet (SWN), while entropy-based filtering is employed to select lexically significant words contributing to sentiment evaluation. The framework is validated across six heterogeneous datasets, demonstrating consistent improvements in F1-score, recall, precision, and accuracy, particularly under conditions of linguistic ambiguity. These results substantiate the effectiveness and domain independence of the proposed approach for sentiment analysis and rating prediction. Furthermore, the framework specifically addresses the challenges posed by the vagueness and uncertainty present in natural language, offering a robust solution for handling ambiguous textual data in real-world applications.