11 Unsupervised Methods

While supervised models often work fine for text classification, one disadvantage is that we need to set specifics for the model. As an alternative, we can not specify anything and have R find out which classifications work. There are various algorithms to do so, of which we here will focus on Latent Dirichlet Allocation (LDA); a ‘seeded’ version of LDA that uses information from other sources to improve the results of the LDA; and a Structural Topic Model.