LLM-assisted literature analysis for plastic upcycling

Area of Science:

• Environmental Science
• Material Science
• Chemistry

Background:

• Plastic waste is a global challenge requiring innovative upcycling solutions.
• The interdisciplinary nature of plastic upcycling research generates a high volume of publications, complicating synthesis.
• Synthesizing insights from extensive plastic upcycling literature is difficult due to its interdisciplinary nature and publication volume.

Purpose of the Study:

• To evaluate the efficacy of large language models (LLMs) in analyzing plastic upcycling research.
• To demonstrate LLMs' capabilities in classifying plastics, identifying upcycling pathways, and visualizing research trends.
• To explore the potential of LLMs in accelerating scientific discovery within the field of plastic upcycling.

Main Methods:

• Analysis of 883 research articles on plastic upcycling.
• Application of large language models (LLMs) for data classification and trend identification.
• Comparative assessment of LLM performance against human expert analysis.

Main Results:

• LLMs demonstrated high efficiency and accuracy in classifying plastics and identifying upcycling pathways.
• LLM performance in defined tasks was comparable to that of human experts.
• LLM-driven analysis was completed significantly faster than traditional methods.

Conclusions:

• LLMs offer a powerful tool for rapid, scalable, and consistent analysis of scientific literature.
• LLM-generated insights can effectively guide future research directions in plastic upcycling.
• A collaborative framework is proposed to optimize LLM integration into scientific workflows for accelerated innovation.