Mathematical modeling by optimization frequently focuses on identifying the extreme values of real-world objectives, such as maximum profit, performance, or yield, and minimum loss, risk, or cost. The standard practice is for a domain expert to manually formulate the models. The mathematical modeling language, akin to programming languages, is precise and unambiguous, making it easy for machines to interpret but challenging for users to master.

The rapid advancement of deep learning technologies has led to large-scale pre-trained language models with remarkable natural language understanding capabilities. By leveraging machine translation techniques, we aim to convert user-friendly natural language into machine-friendly modeling language, ultimately making advanced algorithms and solvers more accessible across various application domains.