UC Berkeley

This dissertation explores the application of data-driven and machine learning-based ap- proaches to address challenges in materials discovery and design. The main objectives in- clude accelerating materials discovery, determining crystal structures, an extracting knowl- edge from research literature. By combining large materials structure databases with fitness metrics, integrating computational methods with experimental data, and utilizing natural language processing and machine learning techniques, this research demonstrates the use of these methods for accelerating progress in the field of materials science and engineering. Overall, the work presented here represents meaningful advances in the state-of-the-art of data-driven methodologies for enhancing our understanding of structure-property relation- ships, expediting materials discovery, and harnessing untapped knowledge from scientific research literature.