Crowdsourced taxonomy verification: A feedback-driven framework for refining knowledge graph relationships via online search interactions

Crowdsourced taxonomy verification: A feedback-driven framework for refining knowledge graph relationships via online search interactions

摘要

一种反馈驱动的验证引擎，通过将未验证的知识图谱边作为假设注入搜索界面，利用用户实时交互信号（如点击、停留）来验证关系准确性。系统平衡探索与利用，自动提升正确边、剪除AI幻觉，无需人工标注，适用于快速变化的商品目录场景。