Low-frequency word perception source code annotation generation method and tool

Abstract

Various embodiments of the disclosure relate to low frequency word aware source code annotation generation and tools. The low-frequency word perception source code annotation generation method comprises the steps that character or character string merging is conducted on annotations of source codes on the basis of a byte pair coding model, so that the annotations in the source codes are segmented; the editing weight is obtained based on the editing distance between a first character or character string and a second character or character string, meanwhile, semantic retrieval is performed on the second character or character string through Faise, the first character or character string is a character or character string of a source code retrieved based on bit parallel computing, and the second character or character string is a character or character string of a source code retrieved based on bit parallel computing. The second character or character string is a to-be-tested character or character string of the source code; and based on the fusion weight, fusing the retrieved source code information into the to-be-tested source code, generating a sequence of sub-words, and merging the sub-words to obtain an annotation of the to-be-tested source code. In this way, combined low-frequency words in code annotations can be effectively predicted, source code annotations can be efficiently generated, and Chinese annotations can be expanded.

Description

technical field [0001] The invention relates to the field of computers, in particular to a method and tool for generating low-frequency word-aware source code annotations. Background technique [0002] The goal of automatic source code annotation is to automatically generate annotations that can accurately describe the source code. These annotations include descriptions in the code and related content in development documents, which are crucial to understanding and maintaining software systems. Using these annotations, developers can understand and maintain software systems more effectively and quickly, free themselves from the tedious task of writing annotations, improve reading speed, quickly and accurately grasp the semantics of source code, quickly locate specific code locations, and judge needs Analyze which parts of your source code in detail to develop more efficiently. [0003] At present, the research on automatic source code annotation is mainly divided into three...

AI Technical Summary

Problems solved by technology

[0009] In addition, the fusion-based annotation generation method has a weak prediction effect on combined low-frequency words, and the performance of the retrieval fusion process is poor, which seriously affects the efficiency of the entire model to generate annotations

[0010] Moreover, the existing methods for automatically generating code comments are only for English comments, and there is a lack of research and exploration related to Chinese comments.

Images