| RAKG is a knowledge graph construction framework that leverages large language models for automated knowledge graph generation. The framework processes documents through sentence segmentation and vectorization, extracts preliminary entities, and performs entity disambiguation and vectorization. The processed entities undergo Corpus Retrospective Retrieval to obtain relevant texts and Graph Structure Retrieval to get related KG. Subsequently, LLM is employed to integrate the retrieved information for constructing relation networks, which are merged for each entity. Finally, the newly built knowledge graph is combined with the original one. |
- Python 3.11
- Conda (recommended)
- Clone the repository:
git clone https://github.com/RAKG/RAKG.git
cd RAKG- Create and activate a conda environment:
conda create -n RAKG python=3.11
conda activate RAKG- Install dependencies:
pip install -r requirements.txtEdit src/config.py to configure your model provider settings:
- For local Ollama: Set
base_urltohttp://localhost:11434/v1/ - For server-based Ollama: Set
base_urltohttp://your_server_ip
Default Ollama model configurations:
- Main model: Qwen2.5-72B, requires good instruction following
- Similarity check model: Qwen2-7B, using smaller model for faster processing
- embedding model: BGE-M3
- Set your OpenAI API key in
OPENAI_API_KEY - Configure model selection:
- Main model: Qwen2.5-72B-Instruct
- Similarity check model: Qwen2.5-14B-Instruct
- Embedding model: BGE-M3
To switch between providers, set USE_OPENAI = True for OpenAI or False for Ollama.
To process text input:
cd examples
python RAKG_example.py --input "your input text" --output result/kg.json --topic "your_topic" --is-textTo process document input:
python RAKG_example.py --input data/MINE.json --output result/kg.jsonTo reproduce the results from the paper:
cd src/construct
python RAKG.pycd src/eval/llm_evalFor evaluation purposes, we recommend using the DeepEval platform. Please refer to the DeepEval documentation for setup and usage instructions.
cd src/eval/MINE_eval
python evaluate_MINE_RAKG.pycd src/eval/ideal_kg_eval
python kg_eval.pyWe welcome contributions! Please read our contributing guidelines before submitting pull requests.
This repo benefits from:
Thanks for these wonderful works.
For any questions or feedback, please:
- Open an issue in the GitHub repository
- Reach out to us at 2212855@mail.nankai.edu.cn