OGA-KG Framework（Quantum-GNN Augmented Knowledge Graph）Tool Dataset

The QGA-KG Framework is a multi-level world model and knowledge graph framework designed for high-complexity scientific research and decision-making scenarios. It organises the following components within a unified system: a four-tier granular lossless knowledge graph, spatially heterogeneous graphs, vector semantic spaces, temporal-causal world models, quantum-enhanced analysis, and neural network inference with application interfaces. Constructed from the bottom up, the framework comprises seven layers: ① The four-tier lossless knowledge graph layer, which maintains high-fidelity semantic and evidential mapping based on multimodal raw data including text, images, tables, and geographic/log records; ② The spatial heterogeneous intelligent graph layer, which integrates entities and relationships from diverse sources, scales, and types into a spatially heterogeneous knowledge structure; ③ A vector and semantic space layer, building upon the graph to establish embedding, similarity retrieval, and Retrieval-Augmented Generation (RAG) mechanisms; ④ Temporal-Causal World Modelling Layer: employs time-augmented world models and causal graphs to model event evolution, strategic pathways, and causal chains; ⑤ Quantum-Augmented Analysis Layer: reserves interfaces for quantum solvers and hybrid pipelines to invoke quantum/quasi-quantum algorithms in combinatorial optimisation and complex game analysis; ⑥ Neural Extraction and Reasoning Layer: Integrates models such as GNNs, RNNs/LSTMs, and Transformers to automate the process from raw data to structured graphs and higher-order reasoning; ⑦ Application and Interface Layer: Provides a unified access point for application scenarios including ‘research co-pilot’ support, policy and commercial decision-making, and VR/AR visualisation.The framework's design objective is to unify traditional knowledge graphs, semantic vector spaces, causal/temporal models, quantum computing, and deep neural networks within a scalable system without compromising underlying evidence granularity. This enables researchers to execute the full chain of processes—‘evidence management, structural modelling, reasoning analysis, and interactive visualisation’—within a single data universe. Resources hosted on Harvard Dataverse—including graph data, vector files, model configurations, and schematic diagrams—serve as the foundational framework for constructing specialised research co-pilots, complex systems decision support platforms, and multimodal teaching/demonstration systems. This enables reproducible graph-world model research across disciplines such as social sciences, international relations, marketing, and security studies.