Learning Physical Common Sense as Knowledge Graph Completion via BERT Data Augmentation and Constrained Tucker Factorization

---

Dependencies

• PyTorch 1.4.0

• OpenKE (Han et al., 2018)

Triple Classification

• Ours in Table 1: python pc_triple_classification.py --add_constraint

• Without data augmentation: python pc_triple_classification.py --data_dir ./data/kge/original/ --add_constraint

• Without constraint: python pc_triple_classification.py

Link Prediction

• Ours in Table 2: python pc_link_prediction.py --add_constraint

• Without data augmentation: python pc_link_prediction.py --data_dir ./data/kge/original/ --add_constraint

• Without constraint: python pc_link_prediction.py

OpenKE

The results of other methods in Table 1 and Table 2:

• TransE

  • python transe_eval.py

  • python openke_triple_classification.py --model transe

• TransD

  • python transd_eval.py

  • python openke_triple_classification.py --model transd

• RESCAL

  • python rescal_eval.py

  • python openke_triple_classification.py --model rescal

• DistMult

  • python distmult_eval.py

  • python openke_triple_classification.py --model distmult

• ComplEx

  • python complex_eval.py

  • python openke_triple_classification.py --model complex

• SimplE

  • python simple_eval.py

  • python openke_triple_classification.py --model simple

• Tucker

  • Triple classification: python tucker_triple_classification.py

  • Link prediction: python tucker_link_prediction.py

Cross Validation

• Triple classification: python cv.py

• Link prediction: python cv_lp.py

Acknowledgements

• The implementation is mainly modified from Balazevic et al., 2019.

• The code of loading physical commonsense data and calculating evaluation metrics is taken from Forbes et al., 2019.

References

• Maxwell Forbes, Ari Holtzman, and Yejin Choi. 2019. Do neural language representations learn physical 457 commonsense? Proceedings of the 41st Annual 458 Conference of the Cognitive Science Society.

• Ivana Balazevic, Carl Allen, and Timothy Hospedales. 2019. TuckER: Tensor factorization for knowledge graph completion. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 5185–5194, Hong Kong, China. Association for Computational Linguistics.

• Xu Han, Shulin Cao, Xin Lv, Yankai Lin, Zhiyuan Liu, Maosong Sun, and Juanzi Li. 2018. OpenKE: An open toolkit for knowledge embedding. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 139–144, Brussels, Belgium. Association for Computational Linguistics.