Code snippet recommendation method based on sequence-to-sequence model

Abstract

The invention discloses a code snippet recommendation method based on a sequence-to-sequence model. The method comprises the steps of collecting and cleaning code snippets and corresponding queries; extracting code information; performing an identifier replacement operation on the code snippets; combining the code snippets after the identifier replacement operation and the corresponding queries into a set of (code, query) pairs, and taking the set as a training set of a query generation model; training a query generation model by using the sequence-to-sequence model; generating a query for thecode snippets of the recommendation code library; inputting a query, and executing similarity calculation between the query and the generated query; and according to a similarity calculation result,constructing a generated query list according to a descending order of similarity scores, and retrieving code snippets corresponding to queries in the list from a recommendation code library to serveas code snippet recommendation results. The method is superior to some existing research works, the code snippet recommendation effect can be effectively improved, and the software development efficiency is improved.

Description

technical field [0001] The invention belongs to the technical field of software engineering, and in particular relates to a method for implementing code segment recommendation by using program static analysis, natural language processing and neural network technology. Background technique [0002] In the current open source software ecological environment, massive codes and various related code description information can be obtained openly and free of charge, and the number and scale of software are rapidly expanding, which also lays the foundation for code reuse. Gabel et al. conducted statistics and research on about 420 million lines of code of more than 6,000 open source projects, and found that software usually lacks uniqueness, that is, program code is often repeated, which also provides theoretical support for code reuse. Effective code reuse can improve developers' development efficiency and reduce development costs. [0003] In the actual software development proc...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the deficiencies in the prior art above, the purpose of the present invention is to provide a method for recommending code fragments based on a sequence-to-sequence model to solve the problem of semantic differences between queries and codes in the prior art; the method of the present invention no longer It directly calculates the similarity between the query and the code, but first trains a query generation model to generate a query for the code fragment, and then, by calculating the similarity between the query input by the user and the generated query corresponding to the code fragment, the retrieval And recommend highly relevant code snippets to developers

Images