Acta Scientiarum Naturalium Universitatis Pekinensis ›› 2023, Vol. 59 ›› Issue (5): 735-746.DOI: 10.13209/j.0479-8023.2023.049

Previous Articles     Next Articles

Knowledge-Driven Interactive Graph Search

LI Yingxue, CHEN Shaohan, ZHENG Weiguo   

  1. School of Data Science, Fudan University, Shanghai 200433
  • Received:2022-08-15 Revised:2022-09-22 Online:2023-09-20 Published:2023-09-18
  • Contact: ZHENG Weiguo, E-mail: zhengweiguo(at)


李映雪, 陈劭涵, 郑卫国   

  1. 复旦大学大数据学院, 上海 200433
  • 通讯作者: 郑卫国, E-mail: zhengweiguo(at)
  • 基金资助:


The existing interactive graph search methods are mainly limited to optimize the annotation cost of single data. To solve this problem, this paper introduces a knowledge-driven method of modeling prior probability information for batch data annotation tasks which are more common in real scenes. This method extracts knowledge between entities of batch data and guides machine algorithms, thus reducing the cost of interactive graph search on the whole. The results of experiments on real datasetsverifies the superiority of proposed algorithm in terms of interaction efficiency compared with existing methods.

Key words: graph search, human-machine interaction, entity knowledge, knowledge-driven


现有的交互式图搜索方法主要局限于优化单一数据的标注成本。为解决这一问题, 针对现实场景中更常出现的批量数据标注任务, 提出一种基于知识驱动建模先验概率信息的方法。利用该方法对批量数据的实体间知识进行提取, 并用于指导机器算法, 可以在整体上降低交互式图搜索的成本。在真实数据集上的实验结果表明, 与现有方法相比, 所提出的算法具有交互效率方面的优势。

关键词: 图搜索, 人机交互, 实体知识, 知识驱动