Engineering intelligent healthcare systems: understanding medical queries with AI and NLP

Abstract Industry 5.0 introduces a human-centered approach where engineering and applied science are combined to create smarter systems that directly improve human well-being. In healthcare, this approach is realized through Healthcare 5.0, which uses Artificial Intelligence (AI) and Natural Language Processing (NLP) to design intelligent platforms that can interpret patient questions and provide accurate responses. This study addresses the engineering challenge of intent classification in medical question-answering systems, an essential step in developing reliable healthcare chatbots and decision-support tools. Using the MedQuad dataset of 14,979 labeled medical questions, we evaluate classical machine learning models such as Logistic Regression, Naive Bayes, Support Vector Machines (SVM), and Random Forest, along with the transformer-based BERT model. Nonetheless, to improve classification under imbalanced data, the Synthetic Minority Oversampling Technique (SMOTE) was applied. In the training phase, the Random Forest model attained 100% accuracy, whereas its inference accuracy on unseen data (without SMOTE) was 80%, demonstrating its effectiveness in generalizing beyond the training set, while other models performed moderately, and BERT required more domain-specific tuning. The findings highlight the contribution of computational engineering methods to healthcare applications and demonstrate how applied AI models can support human-centered solutions at the intersection of engineering and medical sciences.