Large language models can summarize texts on a linguistic level. It is currently being investigated whether they can also summarize scientifically complex texts in a way that lay people can understand them.
Researchers from the Chair of the Fundamentals of Natural Language Processing at the University of Bamberg (“BamNLP”) and the KLARpsy service at the Leibniz Institute for Psychology (ZPID) are investigating this question as part of a joint project.
At klarpsy.de, the freely available service provides plain language summaries of current psychological review articles with meta-analysis to enable individuals with no prior knowledge of psychology to understand and contextualize research findings from psychology. The summaries are written according to a guideline containing 37 criteria derived from research. This ensures a high standard of quality.
Based on these criteria, the project will use large language models to develop a tool that will support KLARpsy authors in writing the summaries. This will enable a more efficient creation of the summaries and thus, a broader topical scope for KLARpsy. To ensure correctness, each article will be manually checked by KLARpsy authors and, if needed, revised.
The cooperation project is funded by the state of Rhineland-Palatinate.