Scholarlink

We present a hierarchical text classification system for web content using the IAB (Interactive Advertising Bureau) taxonomy, with a focus on an orchestrator ensemble module, called MAESTRO (Model Aggregation Ensemble with Soft-weighted voTing for Robust Outcomes) that overcomes the limitations of individual classifiers. The ensemble orchestrator improves performance by selecting the most reliable classifier for each taxonomy category, achieving state-of-the-art results on IAB category classification (Macro-F1 up to 0.81 on Tier-1 categories). We also compare with previous systems in literature and discuss how our approach provides better coverage of the taxonomy and more robust classification across diverse domains. Strengths, limitations, and scalability of the system are analyzed.

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

MAESTRO - Model Aggregation Ensemble with Soft-Weighted VoTing for Robust Outcomes in News Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Giovanni Forlenza,
  • Davide Perfetto,
  • Martina Buttarelli,
  • Luca Carrubbo,
  • Francesco Polese,
  • Valentina Rossi,
  • Maria Angela Pellegrino,
  • Antonio Lieto

摘要

We present a hierarchical text classification system for web content using the IAB (Interactive Advertising Bureau) taxonomy, with a focus on an orchestrator ensemble module, called MAESTRO (Model Aggregation Ensemble with Soft-weighted voTing for Robust Outcomes) that overcomes the limitations of individual classifiers. The ensemble orchestrator improves performance by selecting the most reliable classifier for each taxonomy category, achieving state-of-the-art results on IAB category classification (Macro-F1 up to 0.81 on Tier-1 categories). We also compare with previous systems in literature and discuss how our approach provides better coverage of the taxonomy and more robust classification across diverse domains. Strengths, limitations, and scalability of the system are analyzed.