This is the official implementation for Re-Temp: Relation-Aware Temporal Representation Learning for Temporal Knowledge Graph Completion (EMNLP Findings 2023).
Simply run python main.py -d ICEWS14
Requirments：torch, dgl(cuda version)
| Argument | Default | Description |
|---|---|---|
| --gpu | 0 | Set cuda device |
| --dataset -d | ICEWS14 | Dataset used here, please choose from GDELT, ICEWS05-15, ICEWS14, ICEWS14s, ICEWS18, WIKI |
| --dropout | 0.2 | Dropout rate |
| --n-hidden | 200 | Layer hidden dimension |
| --n-layers | 2 | Number of layers of one GNN |
| --history-len | 3 | history length |
| --lr | 0.001 | learning rate |
| --early_stop | 5 | early stop epochs |
| --easy_copy | 0 | Remove most print results, only keep the final output |
@inproceedings{wang2023re,
title={Re-Temp: Relation-Aware Temporal Representation Learning for Temporal Knowledge Graph Completion},
author={Wang, Kunze and Han, Caren and Poon, Josiah},
booktitle={Findings of the Association for Computational Linguistics: EMNLP 2023},
pages={258--269},
year={2023}
}
Some of the codes are inspired by:
We express gratitude for all the previous contributions in this area.
If you can't replicate the results reported in the paper using the default hyperparameter, please consider updating your torch and dgl package version.