Ontopharmsearch: A Scalable Ontology-Driven Semantic Search Framework For Contextual Real-World Evidence Discovery In Pharmaceutical Research

Abstract

The appearance of real-world evidence (RWE) as an addition to randomized controlled trials has revolutionized pharmaceutical research, but the heterogeneous and unstructured nature of real-world data (RWD) constrains conventional keyword-based retrieval systems. These approaches tend to lack biomedical semantics and contextual relationships, which limit the discovery of evidence. To overcome these shortcomings, this paper presents OntoPharmSearch, a scalable semantic search framework based on ontologies (SNOMED CT, UMLS, MeSH) and transformer-based embeddings (BERT, BioBERT) to ensure precise and context-aware search of RWE. The proposed framework combines metadata ingestion, structural standardization, concept normalization, entity disambiguation, ontology alignment, semantic enrichment, and vector embedding generation with approximate nearest-neighbour retrieval to improve interpretability, scalability and efficiency. Experimental analysis on heterogeneous datasets including EHRs, clinical trial registries, and drug safety databases showed better results, with 96.2% accuracy, 96.8% precision and 95.5% recall. These findings imply that OntoPharmSearch is effective in eliminating noise and irrelevant retrieval as well as in providing comprehensive evidence discovery to pharmaceutical research. By overcoming semantic gaps and allowing transparent query processing, the framework brings a considerable improvement over traditional search systems, which are used to support high-quality, context-sensitive decision-making in healthcare and pharmaceutical fields.