Cloud manufacturing (CMfg) is a recent manufacturing paradigm that aims to provide a networked environment to manufacturing providers and customers. Within this network, manufacturing providers can share resources and collaborate to satisfy customer orders. One of the main challenges in the automation of such sharing is selecting multiple resources that can collectively undertake complex tasks in an optimal way. This problem, often referred to as “service composition and optimal selection (SCOS),” has been studied extensively. However, the main body of the work has focused on developing and implementing cutting-edge algorithms to further optimize the Quality of Service (QoS) of the final composite service. In this paper, we target some of the shortcomings of these studies by first, collecting and utilizing a real-world manufacturing dataset, and second, implementing a bidirectional encoder representation from transformers (BERT)-based model on the aforementioned dataset to form the candidate sets. A well-established metaheuristic (genetic algorithm) is also implemented to form the optimal composite solution.