Research On Semantic Modeling Of Test Data Of Programming Problems

With many researches on programming contest, including developing online-judge platforms and contestant training assistant systems, some imaginations or requirements occurs, which hope machines can automatically help the processing of test data of programming problems, including: automatically validating whether each testcase in a problem fits its description in format and data range; automatically cheching how the boundary values of the input variables are covered by the test data; assisting to generate testcases with complicated logical structure and guarantee their correctness; etc... The foundation of implementing these functionalities is that the machine can understand the essential content of the database.On the other hand, in many programming problems, test data is required to be input in form of plain text. These plain texts don't stand for simple variables, but some objects and their status in the background story of the problem. Test data of different problems describe different objects. Even they describe same of similar objects, authors of different problems may define different format to express the objects; Thus, test data from different problems need be processed with different logics.To make a single program process test data of different problems, this thesis proposes the concept model and format model of test data, which describe test data semantically. The concept model of the test data of a problem describes the ontologies of the objects in the background story, while the format data describes the plain text format to express these objects. Once the author of a problem sets up the concept model and format model, a machine can understand what kind of objects each testcase expresses.This thesis proposes the meta-models of the concept model and format model, which is a high level structure followed by the concept models and format models of each specific problem, although they may differ in each problem. With these two meta-models, only one reasoning program is needed, which can process test data of each problem in a uniform way.The achievement of this thesis can be the foundation of researches on test data processing of programming problems.