In this presentation, we will discuss our recent progress of the application of advanced natural language processing (NLP) techniques and large language models (LLMs) across various scientific disciplines to enhance data analysis and knowledge discovery. Firstly, we introduce the Scientific Sentiment Network (SSNet), a novel sentiment analysis model that effectively categorizes expert opinions in the materials science literature into challenges and opportunities, achieving accuracies of 94% and 92% respectively. This approach helps in understanding prevailing scientific viewpoints and forecasting material innovations. Secondly, we discuss the Structured Information Inference (SII) method, which systematically transforms unstructured experimental data into a structured format, thereby updating an existing perovskite solar cell dataset for improved usability in subsequent analyses. Thirdly, the Automatic Generation of Scientific Question Answers (AGSQA) framework is presented. This framework automates the creation and evaluation of QA pairs from extensive scientific texts, enabling deeper engagement and understanding of complex scientific topics. Finally, we describe the DARWIN series of tailored LLMs, which automate the generation of scientific instructions from textual data, reducing reliance on manual data curation and facilitating seamless integration of scientific knowledge into LLMs. Collectively, these methodologies not only streamline scientific workflows but also enrich the quality and applicability of scientific data in research

Speaker Bio:

Professor Bram Hoex completed both an MSc and PhD degree from Eindhoven University of Technology in the Netherlands in 2003 and 2008, respectively. From 2008 to 2015, he worked at the Solar Energy Research Institute of Singapore (SERIS) at the National University of Singapore (NUS) as a Group Leader and from 2012 also as Director.  In 2015, he joined UNSW, where he is currently a Deputy Head at UNSW’s School of Photovoltaic and Renewable Energy Engineering. His research is multifaceted, focusing on the development and application of nanoscale thin films to enhance renewable energy devices, the reliability of solar cells and modules, the financial and performance modeling of large-scale solar farms, and the application of artificial intelligence in material science. He has published over 250 scientific papers and is best known for his groundbreaking work on aluminum oxide for crystalline silicon surface passivation, which is now the de facto standard for industrial PERC and TOPCon solar cells. His work has received various international recognitions, including the 2008 SolarWorld Junior Einstein and 2016 IEEE PVSC Young Professional awards. Renewable Energy World listed him in the “Solar 40 under 40 list” globally in 2018.