In well-defined domains such as Physics, Mathematics, and Chemistry, solutions to a posed problem can objectively be classified as correct or incorrect. In ill-defined domains such as medicine, the classification of solutions to a patient problem as correct or incorrect is much more complex. Typical tutoring systems accept only a small set of approved solutions for each problem scenario fed to the system. Plausible student solutions that fall outside the scope of this small set of approved solutions are rejected as being incorrect, even though these solutions may be acceptable or close to acceptable. This leads to brittleness in the evaluation of student solutions. This paper describes a tutoring system for medical problem-based learning (PBL), which can accept a wide variety of plausible solutions without placing an extensive burden on knowledge acquisition. A widely available medical knowledge source is deployed as a domain ontology, and concept relationships in the ontology are used to make inferences and expand the space of plausible solutions beyond the scope of solutions explicitly provided to the system. Parent-child relationships are used to infer generalized solutions, whereas relationships of synonymy are used to infer alternate solutions. Evaluations of the system after expanding the solution space indicate accuracy close to that of human experts, who agreed among themselves with Pearson Correlation Coefficient of 0.48 and p < 0.05. The system precision dropped by 32%, while the recall increased by five times. The geometric mean of sensitivity and specificity was increased by 0.33.