This paper presents an algorithm for translating concurrent procedural language programs into nonprocedural, mathematical language programs, called specifications. The goal is to achieve reuse of old existing programs in developing new systems, through having them explained automatically and facilitating their modification. Mathematical languages are widely believed to be superior to procedural languages. Unlike procedural languages, mathematical languages do not have "side effects" and are oblivious to computer concepts. Thus mathematical languages free the user of having to "think like a computer" when developing or modifying a program. Its mathematical semantics make proving software correctness easier and improves software reliability. The specification can then be used to generate automatically highly efficient procedural language programs for computer system. The translation algorithm centers around the difference in the meaning of variables in procedural and mathematical languages. In a procedural language a variable may be assigned many values. In a mathematical language, however, a variable may be assigned only one value. The translation algorithm focuses on renaming variables in a procedural language program so that each variable is assigned only one value. This paper also presents a methodology for proving specification correctness. The idea is based on generating scenarios that define values of variable for an applicable situation and using this to prove the specification satisfy a given requirement. This is contrasted with use of temporal logic for proving correctness of concurrent programs.