Generating Commit Messages from Diffs using Pointer-Generator Network

Liu, Qin; Liu, Zihe; Zhu, Hongming; Fan, Hongfei; Du, Bowen; Yu, Qian

doi:10.1109/msr.2019.00056

Cited by 53 publications

(37 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the task of applying changes to code, we show how to make edit representations more generalized and report better performance compared to other neural models proposed in prior work for this task. In the commit message generation task, our model achieves the same results as neural machine translation models trained specifically for this task [10,14]. This indicates that the constructed embeddings of code changes contain enough information to describe them in natural language.…”

Section: Introductionmentioning

confidence: 63%

“…For example, Jiang et al [10] gather a dataset of Git commits and their messages and train a neural machine translation model with the attention mechanism [1] to generate commit messages from code changes. Liu et al [14] add the copying mechanism to the model of Jiang et al and report better performance on the same dataset. Hoang et al [8] propose a tool called PatchNet to predict stable patches in the Linux kernel.…”

Section: Introductionmentioning

confidence: 99%

“…• Based on the proposed approach, we implement a model called CodeChangeEmbedder. We make all the data and code publicly available, including the replication packages not only for our work but also for other papers that we use as baselines [10,14,29], which makes reproducibility and direct comparison with them easier for other researchers in this area. You can find all the data in the replication package: https://zenodo.org/record/4892522.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Unsupervised Learning of General-Purpose Embeddings for Code Changes

Pravilov¹,

Bogomolov²,

Golubev³

et al. 2021

Preprint

View full text Add to dashboard Cite

A lot of problems in the field of software engineering -bug fixing, commit message generation, etc. -require analyzing not only the code itself but specifically code changes. Applying machine learning models to these tasks requires us to create numerical representations of the changes, i.e. embeddings. Recent studies demonstrate that the best way to obtain these embeddings is to pre-train a deep neural network in an unsupervised manner on a large volume of unlabeled data and then further fine-tune it for a specific task.In this work, we propose an approach for obtaining such embeddings of code changes during pre-training and evaluate them on two different downstream tasks -applying changes to code and commit message generation. The pre-training consists of the model learning to apply the given change (an edit sequence) to the code in a correct way, and therefore requires only the code change itself. To increase the quality of the obtained embeddings, we only consider the changed tokens in the edit sequence. In the task of applying code changes, our model outperforms the model that uses full edit sequences by 5.9 percentage points in accuracy. As for the commit message generation, our model demonstrated the same results as supervised models trained for this specific task, which indicates that it can encode code changes well and can be improved in the future by pre-training on a larger dataset of easily gathered code changes.

show abstract

Section: Introductionmentioning

confidence: 63%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Learning of General-Purpose Embeddings for Code Changes

Pravilov¹,

Bogomolov²,

Golubev³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…There are also more advanced NMT-based commit message generation methods. PtrGNCMsg [9] is a new approach that addresses the out-of-vocabulary (OOV) problem by using pointer-generator network. In this manner, before generating each of the commit message words, PtrGNCMsg uses probabilistic techniques to decide whether the word should be selected from the training vocabulary or copied from the given test commit diff.…”

Section: Neural-machine-translation-based Methodsmentioning

confidence: 99%

On the Relevance of Cross-project Learning with Nearest Neighbours for Commit Message Generation

Etemadi

Monperrus

2020

Proceedings of the IEEE/ACM 42nd International Conference on Software Engineering Workshops

View full text Add to dashboard Cite

Commit messages play an important role in software maintenance and evolution. Nonetheless, developers often do not produce highquality messages. A number of commit message generation methods have been proposed in recent years to address this problem. Some of these methods are based on neural machine translation (NMT) techniques. Studies show that the nearest neighbor algorithm (NNGen) outperforms existing NMT-based methods, although NNGen is simpler and faster than NMT. In this paper, we show that NNGen does not take advantage of cross-project learning in the majority of the cases. We also show that there is an even simpler and faster variation of the existing NNGen method which outperforms it in terms of the BLEU_4 score without using cross-project learning. CCS CONCEPTS • Software and its engineering → Software maintenance tools.

show abstract

“…Furthermore, Jiang et al [15] add filters to filter out the likely poor prediction, and thereby improve the performance of the commit message generation. Liu et al [25] alleviate the out-of-vocabulary problem by adopting the pointer network. Although deep learning is a promising new technique, it is still debatable as to whether this method can be implemented in a way that benefits SE [10,13].…”

Section: Deep Learning In Software Engineeringmentioning

confidence: 99%

SEED: Semantic Graph based Deep detection for type-4 clone

Xue,

Liu,

Zeng

et al. 2021

Preprint

View full text Add to dashboard Cite

Background: Type-4 clones refer to a pair of code snippets with similar functionality but written in different syntax, which challenges the existing code clone detection techniques. Previous studies, however, highly rely on syntactic structures and textual tokens, which cannot precisely represent the semantic information of code and might introduce nonnegligible noise into the detection models. Aims: To overcome these limitations, we explore an effective semantic-based solution for Type-4 clone detection. Additionally, we conducted an empirical study on the characteristics of Type-4 clone pairs. We found that NOT all tokens contain semantics that the Type-4 clone detection required. Operators and API calls emerge as distinct candidates for Type-4 code semantic representation. Method: To bridge this gap, we design a novel semantic graph based deep detection approach, called SEED. For a pair of code snippets, SEED constructs a semantic graph of each code snippet based on intermediate representation to represent the code functionality more precisely compared to the representations based on lexical and syntactic analysis. To accommodate the characteristics of Type-4 clones, a semantic graph is constructed focusing on the operators and API calls instead of all tokens. Then, SEED generates the feature vectors by using the graph deep neural network and performs code clone detection based on the similarity among the vectors. Results: Extensive experiments show that our approach significantly outperforms two baseline approaches over

show abstract

Generating Commit Messages from Diffs using Pointer-Generator Network

Cited by 53 publications

References 25 publications

Unsupervised Learning of General-Purpose Embeddings for Code Changes

Unsupervised Learning of General-Purpose Embeddings for Code Changes

On the Relevance of Cross-project Learning with Nearest Neighbours for Commit Message Generation

SEED: Semantic Graph based Deep detection for type-4 clone

Contact Info

Product

Resources

About