Exploring Examinees’ Responses to Constructed Response Items with a Supervised Topic Model

This study introduces a supervised topic model that incorporates finite-mixture modelling into supervised latent Dirichlet allocation. This model can detect latent groups of participants that have different relationships between their textual responses and associated scores.