Similar Items: Threat Modelling using Domain-Adapted Language Models: Empirical Evaluation and Insights