Program code duplicate checking method based on hierarchical attention neural network

Abstract

The embodiment of the invention discloses a program code duplicate checking method based on a hierarchical attention neural network. The method comprises the following steps: a building step of a codeduplicate checking system, a program code preprocessing step, a building step of a classification model based on a hierarchical attention network, and a program code duplicate checking model trainingevaluation and prediction step based on the hierarchical attention network. By utilizing the embodiment of the invention, a text representation technology of natural language processing can be combined with a program code duplicate checking task, and the program code duplicate checking method which is strong in anti-interference capability, capable of simultaneously extracting program code characteristics and program structure characteristics, capable of processing long program codes and suitable for various high-level languages is provided.

Description

technical field [0001] The invention relates to the field of data analysis and processing of non-natural language texts, and emphatically describes a program code checking method based on a hierarchical attention neural network. Background technique [0002] With the development of the Internet, people gradually rely on the use of search engines to search for answers and solutions to problems from the Internet. On the one hand, it makes it easier for people all over the world to share and obtain knowledge. Used for plagiarism made a breeze. In the field of computer programming, the phenomenon of program plagiarism is very widespread. When we search for a programming question on the Internet, many answers from different sources often appear. However, in these different versions of the code, many source codes are the same, and their core parts are almost identical. [0003] In the field of programming language teaching in colleges and universities, the phenomenon of code pla...

AI Technical Summary

Problems solved by technology

The method based on attribute feature statistics relies on the selection and extraction of keywords, and judges whether a code pair is a pair of plagiarism by calculating the similarity measure of different types of tokens. Although this method has achieved good results in some code duplication check data sets effect, but its robustness is low, it does not take into account the structural characteristics of the program, and cannot cope with plagiarism methods such as adding redundant statements, changing data types, and equivalently replacing control logic structures; the tree-based plagiarism check method takes into account the program code The grammatical features of the program code can be learned by building a natural language-like grammatical analysis tree, but the cost of building a grammatical tree is very high, and it cannot deal with the reordering of statements in the code block and the reordering of operands.

[0004] At present, the methods of code plagiarism are becoming more and more complicated, and the difficulty of students' code homework continues to increase, resulting in an increase in the length of program codes. Traditional code checking methods can no longer meet the needs, and there are only a handful of researches on programming text checking.

Images