Scholarlink

<p>Text classification&#xa0;is a cornerstone of natural language processing and artificial intelligence with applications ranging from sentiment analysis to document categorization. In recent years, the field has witnessed significant advancements due to the emergence of deep learning models. This paper presents four novel deep learning models for text classification, based on Double and Triple hybrid architectures using BERT and DistilBERT. The proposed Triple-BERT and Triple-DistilBERT models achieve 100% accuracy on the Classic3 dataset and 99% on the Classic4 dataset. The Double-BERT model performs best on the IMDB dataset with an accuracy of 93.5%, outperforming baseline BERT by 1.6%. The four methods are evaluated on three well-known datasets.</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

A novel double and triple BERT and distilBERT classification methods

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Amira Samy Talaat

摘要

Text classification is a cornerstone of natural language processing and artificial intelligence with applications ranging from sentiment analysis to document categorization. In recent years, the field has witnessed significant advancements due to the emergence of deep learning models. This paper presents four novel deep learning models for text classification, based on Double and Triple hybrid architectures using BERT and DistilBERT. The proposed Triple-BERT and Triple-DistilBERT models achieve 100% accuracy on the Classic3 dataset and 99% on the Classic4 dataset. The Double-BERT model performs best on the IMDB dataset with an accuracy of 93.5%, outperforming baseline BERT by 1.6%. The four methods are evaluated on three well-known datasets.