Bayesian item response theory and latent dirichlet allocation applied to textual professor performance profiling

The evaluation of teaching and the perception about the academic environment are only obtained through students. Measure the teachers’ performance and subject relevance has an increasing importance for universities and academic institutions since these evaluations are subjective. Thus, a natural question is: how to develope an unbiased measure through subjective answers? This thesis uses several statistical tools in order to measure professor performance and subject relevance. An strategic diagram is constructed to compare these latent variables. Item response theory (IRT) is used to propose a method to measure teachers’ performance and subject relevance, also Latent Dirichlet allocation (LDA) is used to model textual data, and clustered in order to profile professors using students’ comments.