STM is a model that assumes the topics have a structural relationship with covariates such as demographics, sentiment, or other metadata. This model tries to find the underlying topics and their structural relationship with other covariates in the observed documents.