Articles are grouped into thematic clusters using an algorithm known as Gibbs Sampling Dirichlet Multinomial Mixture (GSDMM), which groups together article headlines based on the similarity of their constituent words. The number of clusters is decided by a human (a Tattle team member) after some experimentation, with the aim of producing meaningful results. The algorithm does not generate names for the clusters. We have chosen to leave them unnamed to allow flexible interpretation, but they are numbered for identification. The dashboard design is inspired by LDAvis, a visualisation technique for topic models.