A Semi-Continuous Formulation for Goal-Oriented Reduced-Order Models