Carbohydrates play an immense role in different aspects of life. NMR spectroscopy is the most powerful tool for investigation of these compounds. Nowadays, progress in computational procedures has opened up novel opportunities giving an impulse to the development of new instruments intended to make the research simpler and more efficient. In this paper, we present a new approach for simulating (13)C NMR chemical shifts of carbohydrates. The approach is suitable for any atomic observables, which could be stored in a database. The method is based on sequential generalization of the chemical surroundings of the atom under prediction and heuristic averaging of database data. Unlike existing applications, the generalization scheme is tuned for carbohydrates, including those containing phosphates, amino acids, alditols, and other non-carbohydrate constituents. It was implemented in the Glycan-Optimized Dual Empirical Spectrum Simulation (GODESS) software, which is freely available on the Internet. In the field of carbohydrates, our approach was shown to outperform all other existing methods of NMR spectrum prediction (including quantum-mechanical calculations) in accuracy. Only this approach supports NMR spectrum simulation for a number of structural features in polymeric structures.