NITK SCaLAR Lab at the CLEF 2025 SimpleText Track: Transformer-Based Models for Biomedical Sentence Simplification (Task 1.1)

Abstract

This paper presents the participation of the SCaLAR Lab from the National Institute of Technology Karnataka Surathkal (India) in the CLEF 2025 SimpleText Lab. Biomedical texts are often difficult to understand due to complex vocabulary and sentence structures, which limit access to crucial scientific information for non-expert audiences. Making biomedical literature more accessible, we propose two transformer-based simplification pipelines: one combining BioBERT and BioBART with prompts providing definitions, and another using a fine-tuned GPT-2 Medium model for direct simplification. Our dual approach demonstrates effective reduction of lexical and syntactic complexity while preserving medical accuracy, supporting clearer communication and laying the foundation for future work in multilingual and hybrid simplification systems. © 2025 Copyright for this paper by its authors.