Installation

Build AnyBURL (License: https://web.informatik.uni-mannheim.de/AnyBURL/)

1. Create folder build in folder AnyBURL-RE
2. Compile with javac de/unima/ki/anyburl/*.java -d build
3. Package with jar cfv AnyBURL-RE.jar -C build .

Build SAFRAN

1. Create directory build in folder SAFRAN
2. Download and extract boost 1.76.0 to folder SAFRAN
   1. Windows: https://boostorg.jfrog.io/artifactory/main/release/1.76.0/source/boost_1_76_0.zip
   2. Unix: https://boostorg.jfrog.io/artifactory/main/release/1.76.0/source/boost_1_76_0.tar.gz
3. Have cmake installed (> 9.6.0)
4. Run cmake ../ from newely created folder build
5. Run make

Install openbiolink

pip install torch openbiolink

Training

Run from project root:

python3 train.py

Evaluation

For maximum (default AnyBURL) aggregation approach run:

python3 save_test_submission.py ./dataset/prediction_max

For Non-redundant Noisy-OR (SAFRAN) aggregation approach run:

python3 save_test_submission.py ./dataset/prediction_nrno