The construction industry faces significant supply chain risks, necessitating effective management strategies. This paper presents a novel approach using transformer-based models for Named Entity Recognition (NER) to identify risk-related entities within construction supply chain management (SCRM) from news articles. It specifically explores the efficacy of two transformer models, BERT and DeBERTa, and employs Genetic Algorithms (GAs) for optimizing model hyperparameters. This research underscores the transformative potential of NLP-driven solutions in enhancing SCRM, particularly within the unique context of the Australian construction industry. The findings highlight the importance of precision in entity recognition for effective SCRM and demonstrate the superiority of DeBERTa in precision-focused tasks, making it a promising tool for practitioners in this field.