A Code Description Semantics Vector Based Java Code Search

The goal of code search is to assist developers in searching for code that meets development needs.It is one of the important research areas of software engineering.Traditional code search uses string matching techniques to directly match words in search text with words in code,or to convert search text and code into intermediate representations for matching,ignoring information such as contextual semantics of search text and code.This leads to a lower accuracy of code search.In view of the above shortcomings,this paper proposes a Java code search method based on code description semantic vector?CDSVJCS,Code Description Semantics Vector based Java Code Search?.This method relies on the open source project source code of the developer's active Github project hosting platform and the Q&A data of the Stack Overflow open source community question answering system.Using the mapping between code text description and code block,the code description-code block library is extracted and built.",provides the basic data of the code search;secondly,CDSVJCS uses syntactic analysis and fastText word vector to synthesize the search text and code description into the vector space,and calculate the similarity between the search text and the code description from the perspective of the context.Combining the mapping relationship between the code description and the code block,the initial search result is generated.Finally,based on the initial search result,the classification algorithm based on LSTM-RNN-based similar code detection is used to divide the search result into a code set that conforms to the search requirement and not A code set that meets the search requirements and uses the code set that meets the search requirements as the final search result.The code description text collected in the tutorials website is used as the input text of the search.The algorithm proposed in this paper is tested and compared with the existing code search method.The effectiveness of the CDSVJCS algorithm is verified by manually evaluating the search results.The experimental results show that the algorithm proposed by the text has improved relative to the existing algorithms in terms of ₁F,F₁₀,and PDCG values.