Scholarlink

In the rapidly evolving field of information retrieval, enhancing the quality and safety of online search results is crucial. This research focused on the integration of Generative AI into online search engines, emphasizing toxicity classification to improve the response quality of reasoning engines through toxicity classification. By bridging traditional search engines with advanced reasoning engines, the study demonstrated the potential of generative models to enhance search quality and safety. Using TF-IDF vectorization and evaluating models like Bagging Classifier, AdaBoost Classifier, XGB Classifier, and Radom Forest Classifier, the research achieved high performance, with BaggingClassifier leading with 95% accuracy and an F1-score of 0.94. The future work focuses on implementation of Federated Learning and models like BERT and mBERT to encompass real time data handling, multiple language models and techniques like back translation for data augmentation. The study underscores the improvement of generative AI in creating more intelligent, fair, and user-centric search experiences by effectively identifying and filtering toxic content.

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

Optimizing Reasoning Engines Using TF-IDF Vectorization Approach for Enhanced Toxicity Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Manya Girdhar,
  • Sudhakar Kumar,
  • Sunil K. Singh,
  • Rashmi Pal,
  • Varsha Arya,
  • Razaz Waheeb Attar,
  • Brij B. Gupta,
  • Kwak-Toi Chui

摘要

In the rapidly evolving field of information retrieval, enhancing the quality and safety of online search results is crucial. This research focused on the integration of Generative AI into online search engines, emphasizing toxicity classification to improve the response quality of reasoning engines through toxicity classification. By bridging traditional search engines with advanced reasoning engines, the study demonstrated the potential of generative models to enhance search quality and safety. Using TF-IDF vectorization and evaluating models like Bagging Classifier, AdaBoost Classifier, XGB Classifier, and Radom Forest Classifier, the research achieved high performance, with BaggingClassifier leading with 95% accuracy and an F1-score of 0.94. The future work focuses on implementation of Federated Learning and models like BERT and mBERT to encompass real time data handling, multiple language models and techniques like back translation for data augmentation. The study underscores the improvement of generative AI in creating more intelligent, fair, and user-centric search experiences by effectively identifying and filtering toxic content.