Scholarlink

<p>Overlapping community detection plays a crucial role in analyzing graph data structure. However, the traditional detection process optimizing through multiple iterations often increases the complexity of graph visualization. The rapid development of large language models and prompt engineering has demonstrated the ability to offer the potential to facilitate intelligent visual analytics. In light of this, we design <b>ComVis</b>, a more efficient interactive visualization system that leverages large language models to detect overlapping <Emphasis Type="BoldUnderline">Com</Emphasis>munities <Emphasis Type="BoldUnderline">Vis</Emphasis>ual analysis of graph data, significantly enhancing graph visualization capabilities via a four-level prompt model. By focusing on the complex graph elements and inhomogeneous distribution structures, ComVis can help users understand community encoded attributes, node statistical attributes, link cognitive attributes, and group domain knowledge, thus gaining insight into the community from local to global. Two usage scenarios and a user study demonstrate the effectiveness of ComVis in aiding users in detecting overlapping communities of graph data.</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

ComVis: a four-level prompt model for overlapping community detection with graph visualization

  • 期刊论文

  • 发表日期：

    2025年8月22日

查看全文
  • Xuan Guo,
  • Bohao Li,
  • Huanyu Bi,
  • Juanxiu Yu,
  • Yihong Ren,
  • Junnan Liu,
  • Mingliang Xu

摘要

Overlapping community detection plays a crucial role in analyzing graph data structure. However, the traditional detection process optimizing through multiple iterations often increases the complexity of graph visualization. The rapid development of large language models and prompt engineering has demonstrated the ability to offer the potential to facilitate intelligent visual analytics. In light of this, we design ComVis, a more efficient interactive visualization system that leverages large language models to detect overlapping Communities Visual analysis of graph data, significantly enhancing graph visualization capabilities via a four-level prompt model. By focusing on the complex graph elements and inhomogeneous distribution structures, ComVis can help users understand community encoded attributes, node statistical attributes, link cognitive attributes, and group domain knowledge, thus gaining insight into the community from local to global. Two usage scenarios and a user study demonstrate the effectiveness of ComVis in aiding users in detecting overlapping communities of graph data.