HVRM/H-VRAM

H-VRAM（Hierarchical Value Representation and Alignment Modeling）是一个结构化数据集，用于组织人权和宪法材料，形成一个分层知识图谱。图谱包含两种主要节点类型：ValueConcept（抽象原则、权利和价值声明）和NormativeRule（具体义务、禁止、许可和程序）。节点通过语义关系（如aligns_with、conflicts_with、concretizes）连接，支持跨司法管辖区比较、规范冲突分析、法律推理和图检索研究。数据集包含来自237个源文档的32,063个规范性规则和17,926个价值概念，以及57,867个跨文档关系。数据以JSON格式存储，分为每个文档的分片文件（rules.json和values.json）、跨文档关系文件（cross_relations.jsonl）和整合的层次图文件（hierarchy.json）。H-VRAM (Hierarchical Value Representation and Alignment Modeling) is a structured dataset designed to organize human rights and constitutional materials into a hierarchical knowledge graph. The graph contains two primary node types: ValueConcept (abstract principles, rights, and value declarations) and NormativeRule (specific obligations, prohibitions, permissions, and procedures). Nodes are connected via semantic relationships such as aligns_with, conflicts_with, and concretizes, enabling research on cross-jurisdictional comparison, normative conflict analysis, legal reasoning, and graph retrieval. The dataset comprises 32,063 normative rules and 17,926 value concepts sourced from 237 source documents, along with 57,867 cross-document relationships. All data is stored in JSON format, split into per-document shard files ("rules.json" and "values.json"), a cross-document relationship file ("cross_relations.jsonl"), and an integrated hierarchical graph file ("hierarchy.json").