As one bottleneck for the development of NLP, commonsense reasoning is more and more popular in the NLP community recently. We aim at resolving commonsense reasoning tasks (question answering, dialog generation, semantic role labeling, coreference resolution, etc..) with knowledge graphs (e.g., ASER, ConceptNet, Wordnet, Probase). As this research project is task-oriented, we do not limit the used methods. Both traditional machine learning methods and deep models are encouraged.