This repository provides a framework for scientific discovery through knowledge extraction and multimodal graph-based reasoning. It transforms scientific literature into knowledge graphs, enabling analysis of interdisciplinary relationships, identification of knowledge gaps, and generation of novel material designs. The target audience includes researchers, scientists, and engineers interested in leveraging AI for advanced discovery and innovation.

How It Works

The project constructs an ontological knowledge graph from scientific papers. It employs graph analysis techniques (node degrees, community detection, centrality) and deep node embeddings to uncover hidden connections and facilitate reasoning. Path sampling strategies link dissimilar concepts, enabling the discovery of novel material designs and prediction of material behaviors by identifying isomorphic patterns across diverse domains like science, technology, and art.

Quick Start & Requirements

• Installation: pip install git+https://github.com/lamm-mit/GraphReasoning or pip install -e git+https://github.com/lamm-mit/GraphReasoning.git#egg=GraphReasoning.
• Dependencies: wkhtmltopdf may be required for multi-agent models. For llama.cpp integration, specific build flags are needed.
• Data: Model weights and graph data (e.g., BioGraph.graphml, embeddings) must be downloaded from Hugging Face.
• Resources: Requires Python, networkx, transformers, and potentially llama-cpp-python. GPU acceleration is beneficial for embedding generation and LLM inference.
• Documentation: arXiv Paper

Highlighted Details

• Enables discovery of novel material designs through cross-domain isomorphic mapping (e.g., biology to music, art to materials).
• Provides extensive graph analysis tools: pathfinding, community detection, scale-free network analysis.
• Includes conversational agents for simulating discussions and answering complex questions based on graph data.
• Supports graph generation from text and integration of new knowledge into existing graphs.

Maintenance & Community

The project is associated with Markus J. Buehler at MIT. Further community or maintenance details are not explicitly provided in the README.

Licensing & Compatibility

The repository does not explicitly state a license. Users should verify licensing terms for commercial or closed-source use.

Limitations & Caveats

The README does not specify a license, which may impact adoption. The project relies heavily on large language models for graph generation and reasoning, implying significant computational resource requirements and potential LLM API costs or setup complexities. Some functionalities, like multi-agent models, may require specific system configurations (wkhtmltopdf).