The new advances in Natural Language Processing (NLP) and Machine Learning (ML) have opened up new avenues in Psychometric and Educational Measurement. By merging these fields, researchers can utilize innovative techniques and methods to analyze language-based data efficiently, propose new parameterization of psychometric models, identify relationships between latent cognitive attributes, and analyze item scores with process data. This dissertation makes this explicit by merging NLP and ML with traditional Diagnostic Classification Models (DCMs). It aimed to develop two innovative algorithms using an Additive Bayesian Network (ABN), a machine learning technique that combines Bayesian networks and a Generalized Linear Regression (GLM) to model complex non-linear relationships. The DCM-ABN algorithm is an iterative procedure to explore attribute hierarchies and automatically update attribute mastery classifications and item profiles while improving model fit. The DCM-ABNprocess is an algorithm that fine-tunes attribute mastery profiles based on examinees' thinking and reasoning that was not accounted for by the test scores. Both algorithms were intended to improve the accuracy of attribute mastery classification. This dissertation consists of three studies. Study I is a comprehensive simulation study aimed to evaluate the performance of the DCM-ABN algorithm under various testing conditions. Study II aims to provide empirical evidence of its performance in a dataset with a known attribute hierarchy. Finally, Study III evaluates the applicability of the DCM-ABNprocess in a real data set and includes a Monte Carlo simulation to validate the interpretations of the algorithm. The results were promising, showing that this multidisciplinary fusion has the potential to uncover deeper patterns and connections that were not easily investigated statistically, taking us beyond traditional score-based analyses.
