Extractive Summarization Approaches for Biomedical Literature: A Comparative Analysis

Abstract

Text summarization is one of the key applications of Natural Language Processing that has become more prominent for information condensation. Text summarization minimizes the content of a source document to generate a coherent summary while maintaining vital information from the source. Automatic text summarization has emerged as a key area of research in information extraction, notably in the biomedical field. Keeping pace with the recent research findings on certain issues has become an incredibly difficult challenge for medical specialists and researchers as the number of publications in medical research increases over time. Automatic text summarization approaches for biomedical literature may aid researchers in quickly reviewing research findings by extracting significant information from recent publications. Typical techniques for medical text summarization necessitate domain knowledge. Such systems’ effectiveness depends on resource-intensive clinical domain-specific knowledge bases and preprocessing mechanisms for generating relevant information. This research compares three text summarization approaches word frequency, cosine similarity, and the Luhn algorithm. The comparative results are generated for 20 biomedical articles and evaluated using the Recall-Oriented Understudy for Gisting Evaluation (ROUGE) metric; the outcome demonstrates the best-suited technique for summarization of biomedical literature.