Finding Semantically Similar Clinical Trials using Sentence Embeddings and a Transformer Model

This study presents a semantic similarity finding system for clinical trials using sentence embeddings and cosine similarity. By combining cleaned textual fields into unified vectors and visualising overlaps via word clouds, the model aids in identifying functionally relevant trials. Evaluations using domain-specific vocabulary across diverse therapeutic contexts validate the system’s capability. The approach not only streamlines literature discovery for researchers and clinicians but also offers transparency through interpretable visual feedback, ensuring that retrieval aligns with clinical intent and scientific relevance.