DeepOCL: A deep neural network for Object Constraint Language generation from unrestricted nature language

Object Constraint Language (OCL) is one kind of lightweight formal specification, which widely used for software verification and validation in NASA Management Group projects. Although OCL provides a simple expressive syntax, it hard the developers to write correctly due lacking knowledge mathematical foundations first-order logic, approximately half accurate at first stage development. A deep neural network named DeepOCL proposed, takes unrestricted natural language as inputs automatically outputs best-scored candidates without requiring domain conceptual model that compulsively required existing rule-based generation approaches. To demonstrate validity our proposed approach, ablation experiments were conducted on new sentence-aligned dataset OCLPairs. The show can achieve state art statement generation, scored 74.30 BLEU, greatly outperformed experienced by 35.19%. approach learning generate expression from language. It be further developed CASE tool industry.