High-throughput Biomedical Relation Extraction for Semi-Structured Web Articles Empowered by Large Language Models