0x14d / embedding-operator-knowledge

Embeddings of Knowledge Graphs Containing Procedural Knowledge

Setup

1. Clone the repository with git clone https://github.com/0x14d/embedding-operator-knowledge
2. Open repository with cd embedding-operator-knowledge
3. Python 3.9 is required (2 options)
   • Use the included dev container (requires Docker)
   • Install python manually
4. Create virtual environment (optional)
   1. python -m venv ./venv
   2. source ./venv/bin/activate
5. Install python3.9-dev with sudo apt install python3.9-dev
6. Install requirements with pip install -r requirements.txt

How to:

Generating the embeddings:

In order to train the embeddings & prepare the evaluation either for LinkPrediction (hits@k & AMRI) or Matches (matches@k) and the amount of iterations required execute:

python -m knowledge_infusion.compare_methods.compare_methods --config LinkPrediction --iterations 30

Generating the evaluations

After generating the embeddings execute:

python -m knowledge_infusion.compare_methods.generate_output_files for hits@k and AMRI or python -m knowledge_infusion.compare_methods.generate_output_files for matches@k

This will generate the resulting tables for:

• Each iteration in knowledge_infusion/compare_methods/results/iteration*/_table_format/
• all iterations combined in knowledge_infusion/compare_methods/results/_table_format/