The general problem of self-supervised learning for humor detection is difficult to solve, although a supervised learning route can be utilized to construct a dataset and predict whether there is humor present in a given text.