Although neural network approaches achieve remarkable success on a variety of NLP tasks, many of them struggle to answer questions that require commonsense knowledge. We believe the main reason is the lack of commonsense \mbox{connections} between concepts. To remedy this, we provide a simple and effective method that leverages external commonsense knowledge base such as ConceptNet. We pre-train direct and indirect relational functions between concepts, and show that these pre-trained functions could be easily added to existing neural network models. Results show that incorporating commonsense-based function improves the baseline on three question answering tasks that require commonsense reasoning. Further analysis shows that our system \mbox{discovers} and leverages useful evidence from an external commonsense knowledge base, which is missing in existing neural network models and help derive the correct answer.
@article{arxiv.1809.03568,
title = {Improving Question Answering by Commonsense-Based Pre-Training},
author = {Wanjun Zhong and Duyu Tang and Nan Duan and Ming Zhou and Jiahai Wang and Jian Yin},
journal= {arXiv preprint arXiv:1809.03568},
year = {2019}
}