Recently, attempts have been to request translation services from anonymous crowds. Compared with platform-based language services, services performed by humans have the advantages of flexibility and quality. However, due to the nature of human services the results are not consistent and quality is not assured. This research tries to solve this problem by creating workflows that make collaboration among crowdsourcing workers far more effective and efficient. We model workers and tasks, and calculate the optimal workflow. To confirm the feasibility of this model, we conduct a computational experiment to calculate the best workflow under various parameters. The results are consistent with existing research so the model is useful in understanding crowdsourcing workflows. In addition, a system is developed for realizing crowdsourcing workflows for translation services. Finally, we develop a translation interface that demonstrates the feasibility of the proposed method.