Core-loss spectroscopies using electrons and X-rays, such as electron energy loss near-edge structures (ELNES) and X-ray absorption near-edge structures (XANES), are indispensable tools for materials characterization and development. These techniques provide detailed insights into atomic environments, chemical bonding, and vibrational properties that underpin material functionality. Traditionally, ELNES/XANES analyses have relied on qualitative interpretation or comparisons with reference spectra obtained from experiments and/or simulations. Recent advances in data-driven approaches, however, have enabled more quantitative and predictive use of these spectra. This review highlights newly developed data-driven methodologies that extend far beyond conventional ELNES/XANES analysis. These approaches accelerate ELNES/XANES simulations, enable the extraction of radial distribution functions, and quantify multiple material properties directly from spectral data. To enhance the interpretability of machine learning (ML) predictions, sensitivity analysis is employed to elucidate the relationships between specific spectral features and target properties. The rapid growth of open materials databases, coupled with increasingly powerful ML models, has further fueled these developments. Together, these advances would point to a future in which automated, interpretable and scalable spectroscopy serves as a central driver for deeper understandings and accelerated materials discovery.
Teruyasu Mizoguchi (Thu,) studied this question.