paperbot · PL 论文追踪

全部论文 待读 收藏

P2 Rank Learning-Based Code Readability Assessment with Siamese Neural Networks

基于排序学习的孪生神经网络代码可读性评估
ASE Proceedings of the 37th IEEE/ACM International Conference on Automated Software Engineering2022匹配度 2/10 · 该论文将机器学习用于代码可读性评估，属于ML在软件工程/程序语言中的一般应用，但与用户关注的静态分析、Agentic程序、LLM/Agent改进程序分析、浮点程序分析及贝叶斯程序分析等核心方向没有直接交集；仅标签“机器学习与程序语言”存在弱关联。引用 0 inbox
Qing Mi
机器学习与程序语言
crossref: 2026-09-12

AI 速览

一句话：将代码可读性评估建模为排序学习任务，提出基于孪生神经网络的成对排序模型。

代码可读性自动评估是软件工程中较新的挑战。该文提出将代码可读性评估视为 learning-to-rank 任务，设计基于孪生神经网络的成对排序模型：输入一对代码，输出二者的可读性排序。作者在三个公开数据集上评估该模型，取得 83.5% 准确率、86.1% 精确率、81.6% 召回率、83.6% F-measure 和 83.4% AUC。

问题
如何自动、准确地评估代码可读性，并将其从回归或分类任务转化为排序问题。
方法
用孪生神经网络构建成对排序模型，对输入的代码对进行比较并输出可读性排序。
评估
在三个公开数据集上评估，报告准确率、精确率、召回率、F-measure 与 AUC 等指标。
原文摘要（Abstract）

Automatically assessing code readability is a relatively new challenge that has attracted growing attention from the software engineering community. In this paper, we outline the idea to regard code readability assessment as a learning-to-rank task. Specifically, we design a pairwise ranking model with siamese neural networks, which takes as input a code pair and outputs their readability ranking order. We have evaluated our approach on three publicly available datasets. The result is promising, with an accuracy of 83.5%, a precision of 86.1%, a recall of 81.6%, an F-measure of 83.6% and an AUC of 83.4%.

链接与引用

DOI 原文 ·

BibTeX 
@inproceedings{paperbot7211,
  title = {Rank Learning-Based Code Readability Assessment with Siamese Neural Networks},
  author = {Qing Mi},
  booktitle = {ASE},
  year = {2022},
  doi = {10.1145/3551349.3560440}
}