Topic modeling has become a widely used approach when analyzing constructed responses in both educational and psychological research. The Latent Dirichlet Allocation (LDA) model is one popular model that summarizes information succinctly from a corpus of text with a set of topics. Often, constructed responses and survey data are collected simultaneously and can potentially measure the same latent construct. Researchers have proposed extensions to the LDA model, such as the structural topic model, which predicts the topic space with the observed covariates. We propose to jointly model the topic space with latent traits used in Item Response Theory (IRT), where the link function between the topic proportions within each document follows a nominal response model. This perspective allows researchers to quantify the added value of text when measuring a latent construct with survey responses and potentially change the interpretation of the latent construct by using multiple data sources. A brief simulation study and applied example will be provided as a proof of concept.