Abstract
Nuclear magnetic resonance (NMR) spectroscopy is a key technique for the determination of molecular structure, but accurate chemical shift prediction remains challenging due to data limitations and the high computational cost of quantum mechanical methods, such as density functional theory (DFT) and gauge-including atomic orbital (GIAO). Recent advances in artificial intelligence (AI) and machine learning (ML) provide fast and scalable alternatives with near DFT accuracy. The field has evolved from descriptor-based models and neural networks (NNs) toward deep learning (DL) approaches, particularly graph neural networks (GNNs) and transformers, which learn molecular representations directly from 2D and three-dimensional (3D) structures. These models have been applied to multiple nuclei and complex molecular systems, with hybrid quantum mechanics (QM)-ML and solvent-aware methods improving accuracy and realism. Despite progress, challenges remain in data quality, generalization to new chemical spaces, and uncertainty quantification. This review summarizes recent methods, datasets, benchmarking practices, and broader AI applications in NMR analysis and spectral processing.