Skip to content

Bachelor degree research on heterogeneous information network alignment.

Notifications You must be signed in to change notification settings

1049451037/HIN-Align

Repository files navigation

HIN-Align

Bachelor degree research on heterogeneous information network alignment. Extended work of GCN-Align.

Datasets are from JAPE and IONE.

Environment

  • python>=3.5
  • tensorflow>=1.10.1
  • scipy>=1.1.0
  • networkx>=2.2

Test

unzip dbp15k.zip
chmod +x test.sh
./test.sh

The pre-trained results are in the res/ folder. If you don't want to train by yourself, just see the files in it.

For social network, run:

python train_sn.py --seed 5

For automatically train weights:

python train_auto.py

Citation

Please politely cite our work as follows:

Zhichun Wang, Qingsong Lv, Xiaohan Lan, Yu Zhang. Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks. In: EMNLP 2018.

TODO

  • Change a_ij to sigmoid(a_ij)
  • Combine with TransE (KG) or DeepWalk (SN)
  • Combine with MT
  • Social Network Alignment
  • Iterative or Bootstrapping
  • Use faiss to improve evaluation speed
  • Dimension Reduction or other ways of combination
  • Automatic training for hybrid weights
  • Batched training for GCN
  • Try other GNN models

About

Bachelor degree research on heterogeneous information network alignment.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published