Generating class name in sequential manner using convolution attention neural network
作者：
Highlights：
• We have proposed a novel approach for class name prediction in tokenized manner.
• It uses Copy Convolutional Neural Network to capture the semantics of the class’.
• We also created a new dataset of Python language classes.
• We evaluated the approach on a benchmark dataset and a newly created dataset.
• Results indicate that the proposed approach can predict accurate class names tokens.
摘要
•We have proposed a novel approach for class name prediction in tokenized manner.•It uses Copy Convolutional Neural Network to capture the semantics of the class’.•We also created a new dataset of Python language classes.•We evaluated the approach on a benchmark dataset and a newly created dataset.•Results indicate that the proposed approach can predict accurate class names tokens.
论文关键词：Class name,Source code,Graph embeddings,Recommendation,Convolution network,Name prediction
论文评审过程：Received 5 October 2020, Revised 15 September 2021, Accepted 7 March 2022, Available online 23 March 2022, Version of Record 28 March 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.116854
