These functions help users to find the optimal number of topics for LDA.
divergence(x)a LDA model fitted by textmodel_seededlda() or textmodel_lda()
divergence() computes the average Kullback–Leibler distance between all the
pairs of topic vectors in x$phi. The divergence score maximizes when the
chosen number of topic k is optimal (Deveaud et al., 2014).
Deveaud, Romain et al. (2014). "Accurate and Effective Latent Concept Modeling for Ad Hoc Information Retrieval". doi:10.3166/DN.17.1.61-84. Document Numérique.