Improving scholarship accessibility with reinforcement learning

Introduction. A vast amount of scholarly work is published daily, yet much of it remains inaccessible to the general public due to dense jargon and complex language. We introduce a reinforcement learning approach that fine-tunes a language model to rewrite scholarly abstracts into more comprehensible versions. Method. Our approach utilises a carefully balanced combination of word- and sentence-level accessibility rewards to guide the language model in substituting technical terms with more accessible alternatives, a task which models supervised fine-tuned or guided by conventional readability measures struggle to accomplish. Analysis. We evaluate our model’s performance through readability metrics, factual accuracy assessments and language quality measurements, comparing results against supervised fine-tuning baselines. Results. Our best model adjusts the readability level of scholarly abstracts by approximately six US grade levels—in other words, from a postgraduate to a high school level. This translates to roughly a 90% relative improvement over the supervised fine-tuning baseline, while maintaining factual accuracy and high-quality language. Conclusions. We envision our work as a step toward bridging the gap between scholarly research and the general public, particularly younger readers, and those without a college degree.