BioAbbreviate: A Biomedical Dataset for Abbreviation Expansion and Disambiguation

The BioAbbreviate dataset addresses the challenges of abbreviation expansion and disambiguation in biomedical texts, a critical task for improving health literacy and text simplification. Abbreviations in medical literature often introduce ambiguity, translation issues, and information loss, complicating comprehension for non-experts. This paper presents a large-scale dataset curated from open-access biomedical repositories, including PubMed, PLOS, and PMC Europe, leveraging automated extraction and validation techniques to capture diverse abbreviation-expansion pairs. The dataset supports the development of robust NLP systems for abbreviation detection and expansion, with evaluations demonstrating the superior performance of domainspecific models like BioBERT. Key contributions include scalable data collection, pattern-based abbreviation identification, and contextual disambiguation strategies. The dataset aims to enhance readability, support downstream NLP tasks, and bridge gaps in biomedical text processing