Scholarlink

<p>This study evaluates the performance of various embedding techniques combined with convolutional neural networks (CNNs) for sentiment analysis on two datasets: Amazon product reviews and IMDB movie reviews. Specifically, we assess three embedding approaches: GloVe, FastText, and Word2Vec. Our experiments reveal that the CNN + FastText configuration achieved the highest accuracy of 97.29% on the IMDB dataset, demonstrating its effectiveness in capturing semantic nuances through subword information. In contrast, the CNN + Word2Vec approach consistently underperformed, particularly on the Amazon dataset, with an accuracy of 92.06%. GloVe showed competitive results, in IMDB sentiment analysis, achieving an accuracy of 96.78%. The findings indicate that embedding choice significantly impacts model performance, highlighting the potential of FastText in sentiment analysis tasks.</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

Harnessing CNNs and Embedding Techniques for Enhanced Sentiment Classification

  • 期刊论文

  • 发表日期：

    2025年8月13日

查看全文
  • Dharmendra Dangi,
  • Abhay Sharma

摘要

This study evaluates the performance of various embedding techniques combined with convolutional neural networks (CNNs) for sentiment analysis on two datasets: Amazon product reviews and IMDB movie reviews. Specifically, we assess three embedding approaches: GloVe, FastText, and Word2Vec. Our experiments reveal that the CNN + FastText configuration achieved the highest accuracy of 97.29% on the IMDB dataset, demonstrating its effectiveness in capturing semantic nuances through subword information. In contrast, the CNN + Word2Vec approach consistently underperformed, particularly on the Amazon dataset, with an accuracy of 92.06%. GloVe showed competitive results, in IMDB sentiment analysis, achieving an accuracy of 96.78%. The findings indicate that embedding choice significantly impacts model performance, highlighting the potential of FastText in sentiment analysis tasks.