Source code sematic extraction is a key step in code intelligence tasks.However,the exist-ing code analysis efforts do not extract the semantic information of source code sufficiently.To solve this problem,a new semantic embedding model of source code based on information enhancement is proposed,namely Aug_Src.Information enhancement objects are source code sequences and abstract syntax trees(AST)for source code parsing.The experimental results show that the METEOR metric value of the code summarization generation method based on Aug_Src is 11.5%higher than that of the standard sequence-to-sequence model.,which verifies the feasibility and validity of Aug_Src.
万方数据
维普
