Understanding and Improving Knowledge Graph Embedding for Entity Alignment

Introduction

This repository is the official implementation of Understanding and Improving Knowledge Graph Embedding for Entity Alignment, ICML 2022.

Please see the paper A Benchmarking Study of Embedding-based Entity Alignment for Knowledge Graphs (VLDB 2020) for dataset details.

Dependencies

This project is based on OpenEA. We did not add additional packages compared with the original OpenEA project.

Example

We provide an example of jupyter notebook.

Quick Start

1. Creat a new conda env and install packages

conda create -n openea python=3.6
conda activate openea
conda install tensorflow-gpu==1.8
conda install -c conda-forge graph-tool==2.29
conda install -c conda-forge python-igraph
pip install -r requirement.txt

2. Install the local package

pip install -e .

3. Use the same scripts in OpenEA to run a model with NeoEA:

python main_from_args.py ./args/sea_args_15K.json D_W_15K_V1 721_5fold/1/

The Code Location of NeoEA

1. Neural ontology and neural axioms

./src/openea/approaches/neural_ontology.py

2. Baselines

We slightly modified the source code of the baselines to inject neural ontology into them:

BootEA: ./src/openea/approaches/bootea.py

SEA: ./src/openea/approaches/sea.py

RSN: ./src/openea/approaches/rsn4ea.py

RDGCN: ./src/openea/approaches/rdgcn.py

3. Parameter settings

We added NeoEA hyper-parameters to the original setting files:

BootEA: ./run/args/bootea_args_15K.json

SEA: ./run/args/sea_args_15K.json

RSN: ./run/args/rsn4ea_args_15K.json

RDGCN: ./run/args/rdgcn_args_15K.json

guolingbing/NeoEA