Identification of Emerging Trends Utilizing Large Language Models

Early identification of emerging technologies, often indicated by weak signals, faces key challenges in terms of reliance on domain specialists and limited scalability. To enable timely and reliable detection of emerging signals, a comprehensive examination of relevant scientific and technological (S&T) trends is essential. This paper addresses these challenges by utilizing advanced natural language processing techniques and large language models to create a quantitative driven foresight framework for identifying emerging weak signals in scientific data. Weak signals are explored within the context of water networks by examining approximately 90,000 scientific publications and leveraging two key indicators of emergence: intensity and diffusion. The methodology presented here streamlines the processing of extensive datasets, yielding actionable insights and enhancing foresight capabilities. The proposed framework has the potential to aid strategic planners and domain experts in improving their ability to identify and monitor emerging trends in science and technology.