XinyaoShen / KEDY

KEDY

• Resources for our WSDM 2022 paper: "[Diversified Query Generation Guided by Knowledge Graph]"
• Paper link: https://dl.acm.org/doi/abs/10.1145/3488560.3498431

Dependencies

• CUDA > 11
• Prepare requirements: pip3 install -r requirements.txt.

Training KEDY

1 Seq2Seq

• python make_dataset.py -config config/demo/demo-prep.yml: convert raw corpus to dealable datasets.

• python preprocess.py -config config/demo/demo-prep.yml: opennmt style preprocess.

• python train.py -config config/demo/demo-train.yml

• python translate.py -config config/demo/demo-transl.yml -model exp/demo/models/DEMO_step_0.pt

2 Graph

• Use my_feature_extractor.py and ccig.py to build graph.
• We can do train and inference with train.py.
• For evaluation, Use multi-bleu.perl.