Relationship discovery between two entities is a problem that has to be addressed when constructing a Knowledge Base (KB). A solution to this problem is important because the KB built from the discovered relations can play a key role in down stream tasks, such as analogical reasoning. An example of this kind of reasoning is whether a dog desires cake: a dog is an animal, cake is food, animals desire food, therefore a dog desires cake. We constructed a system that is trained on a commonsense KB and whose inputs are pairs of concepts and its outputs are the strength of commonsense assertions between the concepts. Our approach is unique because it can handle out of vocabulary entities and can generalize commonsense to out of knowledge concepts. We utilize the system to be able to infer the answer for out of knowledge assertions such as the aforementioned whether a dog desires cake.