This paper provides a comprehensive review of the current research on deep learning-based code generation methods. The authors classify the existing deep learning-based code generation methods into three categories: methods based on code features, methods incorporated with retrieval, and methods incorporated with post-processing.
The first category of methods uses deep learning algorithms to generate code based on code features. These methods typically employ sequence-to-sequence models to learn the correspondence between natural language descriptions and code representations from training data.
The second and third categories of methods build upon the first category by incorporating additional components to further improve the performance of code generation. The methods incorporated with retrieval leverage external knowledge bases to enhance the code generation models, while the methods incorporated with post-processing focus on improving the quality and correctness of the generated code.
The paper also summarizes and analyzes the commonly used datasets and evaluation metrics in the existing code generation research. Finally, the authors discuss the overall progress of deep learning-based code generation and provide insights into future research directions that are worth exploring.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Zezhou Yang,... at arxiv.org 04-19-2024
https://arxiv.org/pdf/2303.01056.pdfDeeper Inquiries