Code generation method based on abstract syntax tree structure information enhancement

Abstract

The invention discloses a code generation method based on abstract syntax tree structure information enhancement and a medium wherein the method comprises: obtaining manually annotated code generation data wherein the code generation data comprises a natural language sentence and a corresponding target code; analyzing the target code by adopting an analyzer to obtain an abstract syntax tree corresponding to the target code so as to obtain a prediction sequence of the abstract syntax tree according to the abstract syntax tree, and taking the natural language sentence and the prediction sequence of the abstract syntax tree as training data; a code generation model is established, the code generation model is trained by using the training data, and the code generation model comprises an encoder, a decoder, a historical information enhancement module and a future information enhancement module; and inputting the obtained to-be-processed natural language sentence into the trained code generation model for conversion so as to generate a target code and an abstract syntax tree corresponding to the to-be-processed natural language sentence, thereby reducing the prediction error of the model.

Description

technical field [0001] The present invention relates to the technical field of machine translation, in particular to a method for generating codes based on abstract syntax tree structure information enhancement and a computer-readable storage medium. Background technique [0002] In related technologies, the existing code generation method converts the input source language into the abstract syntax tree of the target code through the model under the framework of the neural network encoder-decoder. Compared with the sequence-to-sequence model, the model can be better The syntax of the code is guaranteed to be correct; however, the model usually only considers the information of the parent node and the previous predicted node when generating the current node of the abstract syntax tree. This method does not perceive the structure of the entire abstract syntax tree and strengthens the model decoding. The machine's perception of the abstract syntax tree leads to poor prediction ...

AI Technical Summary

Problems solved by technology

[0002] In related technologies, the existing code generation method converts the input source language into the abstract syntax tree of the target code through the model under the framework of the neural network encoder-decoder. Compared with the sequence-to-sequence model, the model can be better The syntax of the code is guaranteed to be correct; however, the model usually only considers the information of the parent node and the previous predicted node when generating the current node of the abstract syntax tree. This method does not perceive the structure of the entire abstract syntax tree and strengthens the model decoding. The machine's perception of the abstract syntax tree, resulting in poor prediction performance of the model

Images