The latent topic block model for the co-clustering of textual interaction data

Textual interaction data involving two disjoint sets of individuals/objects are considered. An example of such data is given by the reviews on web platforms (e.g. Amazon, TripAdvisor, etc.) where buyers comment on products/services they bought. A new generative model, the latent topic block model (LTBM), is developed along with an inference algorithm to simultaneously partition the elements of each set, accounting for the textual information. The estimation of the model parameters is performed via a variational version of the expectation maximization (EM) algorithm. A model selection criterion is formally obtained to estimate the number of partitions. Numerical experiments on simulated data are carried out to highlight the main features of the estimation procedure. Two real-world datasets are finally employed to show the usefulness of the proposed approach.