KIT researchers use AI to analyze papers in materials science journals to identify new research trends and innovative ideas. (Illustration: Krot_Studio – Stock.adobe.com/edited)
The number of scientific publications is increasing so rapidly that scientists are no longer able to keep track of them all, even in their own field of research. In the current study, researchers at Karlsruhe Institute of Technology (KIT), in collaboration with scientific partners, have shown how this wealth of information can yield new research ideas. They used artificial intelligence (AI) to systematically analyze materials science publications and identify potential new research avenues. Their results were published in Nature Machine Intelligence. (DOI: 10.1038/s42256-026-01206-y)
Materials are the basis for many technologies, including batteries, solar cells, electronic components, and medical applications. For this reason, materials science has become an interdisciplinary field with a strong influence on many fields of research and technology and a correspondingly large number of research papers. However, the findings described in these papers are only useful if relevant trends and relationships can be identified. Against this background, the authors of this study considered ways to systematically analyze scientific papers. “Our goal is to support the creative thought process of researchers by highlighting new avenues of research and opportunities for interdisciplinary collaboration,” said Professor Pascal Friederich of KIT’s Institute of Nanotechnology.
Combining large-scale language models and machine learning
In their project, the researchers combined large-scale language models (LLM) and machine learning (LM) techniques. The LLM begins by identifying key terms and scientific concepts within journal articles. This information is used to generate a concept graph, a knowledge network where each keyword forms a node. The second machine learning model connects nodes when terms are mentioned together particularly often in scientific papers.
“For example, if our LLM observed that terms like ‘perovskite’ and ‘solar cell’ appeared together more often, new links would be drawn in the concept graph,” said Thomas Marwitz, a computer science student at KIT and lead author of the study. “The ML model then analyzes trends in these associations and predicts which combinations of scientific concepts are likely to become more important over the next two to three years.” The ML model does this by analyzing how the links between terms change over the years. If certain concepts are linked with increasing frequency, this may indicate that a new field of research is developing. Conversely, a decrease in the number of links may indicate that a particular topic is losing relevance.
For materials science research. (Illustration: Thomas Marwitz, KIT)
Ideas for new research areas
The results of the analysis allow researchers to focus on a combination of topics that have previously received little attention. Interviews with experts revealed that they actually think some of the suggestions generated by AI are innovative and promising. “We are not trying to replace researchers,” Friederich stressed. “Our discoveries are not invention machines; they are analytical tools that help us more effectively identify new ideas and collaboration opportunities. Our aim is to provide targeted support for scientific creativity.”
This study shows that AI can be used to systematically analyze large amounts of scientific literature. This approach can also help uncover emerging research trends in other scientific fields.
original publication
Thomas Marwitz, Alexander Colsmann, Ben Breitung, Christoph Brabec, Christoph Kirchlechner, Eva Blasco, Gabriel Cadilha Marques, Horst Hahn, Michael Hirtz, Pavel A. Levkin, Yolita M. Eggeler, Tobias Schlöder, Pascal Friederich: Predicting new research directions in materials science using large-scale language models and concept graphs. Nature Machine Intelligence, 2026. DOI 10.1038/s42256-026-01206-y.
Detailed information
