Comparative Retrieval Outputs of Lexical, TF-IDF, and Semantic Search Methods in a Fixed ACLS Corpus

This dataset contains the ranked retrieval outputs generated for three multi-term clinical queries executed under three retrieval paradigms—lexical (bag-of-words), weighted lexical (TF-IDF), and LLM-enhanced semantic search—within a fixed corpus of 424 open-access biomedical articles on Advanced Cardiac Life Support (ACLS). Each CSV file records ranked results and relevance scores produced by the corresponding retrieval pipeline: query_1_multi_term_top10_full_ranking.csv, query_2_multi_term_top10_full_ranking.csv, query_3_multi_term_top10_full_ranking.csv contain the top-10 ranked outputs for each retrieval method. query_1_multi_term_search_results.csv, query_2_multi_term_search_results.csv, query_3_multi_term_search_results.csv contain the full ranked lists across the entire corpus. All runs were conducted using a controlled Python environment to ensure reproducibility. The dataset supports the doctoral study “LLM-Enhanced vs Lexical Search in Biomedical Literature Retrieval” and illustrates how retrieval paradigm choice—lexical, weighted lexical, or semantic—alters which studies are surfaced in medical information search. The corpus was derived from PubMed Central open-access articles, and all processing conformed to FAIR data and PRISMA-S transparency principles.