THU-KEG / DacKGR

Source codes and datasets for EMNLP 2020 paper "Dynamic Anticipation and Completion for Multi-Hop Reasoning over Sparse Knowledge Graph"

DacKGR

Source codes and datasets for EMNLP 2020 paper Dynamic Anticipation and Completion for Multi-Hop Reasoning over Sparse Knowledge Graph

Requirements

• python3 (tested on 3.6.6)
• pytorch (tested on 1.5.0)

Data Preparation

Unpack the data files

unzip data.zip

and there will be five datasets under folder data.

# dataset FB15K-237-10%
data/FB15K-237-10

# dataset FB15K-237-20%
data/FB15K-237-20

# dataset FB15K-237-50%
data/FB15K-237-50

# dataset NELL23K
data/NELL23K

# dataset WD-singer
data/WD-singer

Data Processing

./experiment.sh configs/<dataset>.sh --process_data <gpu-ID>

dataset is the name of datasets. In our experiments, dataset could be fb15k-237-10, fb15k-237-20, fb15k-237-50, nell23k and wd-singer. <gpu-ID> is a non-negative integer number representing the GPU index.

Pretrain Knowledge Graph Embedding

./experiment-emb.sh configs/<dataset>-<model>.sh --train <gpu-ID>

dataset is the name of datasets and model is the name of knowledge graph embedding model. In our experiments, dataset could be fb15k-237-10, fb15k-237-20, fb15k-237-50, nell23k and wd-singer, model could be conve. <gpu-ID> is a non-negative integer number representing the GPU index.

Train

# take FB15K-237-20% for example
./experiment-rs.sh configs/fb15k-237-20-rs.sh --train <gpu-ID> 

Test

# take FB15K-237-20% for example
./experiment-rs.sh configs/fb15k-237-20-rs.sh --inference <gpu-ID> 

Cite

If you use the code, please cite this paper:

Xin Lv, Xu Han, Lei Hou, Juanzi Li, Zhiyuan Liu, Wei Zhang, Yichi Zhang, Hao Kong, Suhui Wu. Dynamic Anticipation and Completion for Multi-Hop Reasoning over Sparse Knowledge Graph. The Conference on Empirical Methods in Natural Language Processing (EMNLP 2020).