Automatically Extracting Acronyms from Biomedical Text

Acronyms are widely used in biomedical literature. Knowledge of acronyms and their corresponding long forms is a fundamental pre-processing step in the mining of information from biomedical text. In this paper, we present a set of heuristics for finding acronyms and their long forms in biomedical texts. Our accurate and scalable heuristics allowed us to determine over 14,000 acronyms/long form pairs in 6 million English language PubMed abstracts. The overall precision of our term finding heuristics was 97.9%. Our heuristics found about 12,000 novel acronyms that were not in the SPECIALIST lexicon, potentially increasing the SPECIALIST lexicon acronym list by 125%. Our stemmed down acronym list is at http://www.computationalbiology.info/.