Policy and regulatory documents are typically lengthy and hierarchically complex, making them difficult to interpret even for domain experts. Mind maps provide an intuitive way to summarize key concepts and reveal document structure, but existing automatic mind-mapping methods often fail to capture the implicit hierarchical organization and domain-specific semantics of regulatory texts. To address this issue, we propose VisHSEMM, a visual analytics framework for hierarchical structure-enhanced mind mapping. The framework integrates automated hierarchical structure extraction, LLM-based quantitative evaluation, and interactive human-in-the-loop refinement to support the construction, verification, and improvement of regulatory mind maps. In addition, we introduce a quantitative evaluation method based on three dimensions—Concept Identification, Link Construction, and Hierarchy Establishment. A user study and expert interviews demonstrate the usability and effectiveness of the proposed framework.
Liu et al. (Fri,) studied this question.