- KagNet: Knowledge-Aware Graph Networks for Commonsense Reasoning [EMNNLP 2019] Bill Yuchen Lin, Xinyue Chen, Jamin Chen, Xiang Ren.
- Pre-trained model have latent ability in common sense task.
- using GCN -> find best path -> LSTM -> Hierarchical Path-base Attention
- concate the pertained representation.
- How Large Are Lions? Inducing Distributions over Quantitative Attributes [ACL 2019] Yanai Elazar, Abhijit Mahabal, Deepak Ramachandran, Tania Bedrax-Weiss, Dan Roth.
- Rule-base Quantitative Detection, but good than SOTA.