Java Code Clone Detection by Exploiting Semantic and Syntax Information From Intermediate Code-Based Graph

Yuan, Dawei; Fang, Sen; Zhang, Tao; Xu, Zhou; Wang, Liu

doi:10.1109/tr.2022.3176922

Cited by 8 publications

(7 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the above advantages, more and more LLMs are proposed for various software engineering tasks [1], [13], [18]. Similar to LLMs for natural languages, these models are first pretrained on massive software engineering data, e.g., source code collected from GitHub, then further fine-tuned for specific downstream tasks, like code search [19], clone code detection [20], code summarization [21], code optimization [22], and program repair [23]. Due to the powerful representation learning ability of LLMs, they achieve state-of-the-art results on these tasks.…”

Section: Large Language Modelmentioning

confidence: 99%

Enhancing Bug Report Summaries Through Knowledge-Specific and Contrastive Learning Pre-Training

Shao,

Xiang

2024

IEEE Access

View full text Add to dashboard Cite

Bug reports are crucial in software maintenance, with concise summaries significantly enhancing the efficiency of bug triagers and ultimately contributing to the development of high-quality software products. Contemporary methods for automatic bug report summarization primarily utilize neural networks' robust learning capabilities. However, these approaches often produce suboptimal summaries due to two primary limitations: 1) the difficulty in assimilating the domain-specific knowledge inherent in bug reports, and 2) the limitations of purely supervised learning in comprehending the comprehensive context of bug reports. To address the above two problems, in this paper, we propose a new approach for bug report summarization, namely KSCLP, which leverages large language models and domain-specific pre-training strategies, i.e., Knowledge-Specific and Contrastive Learning Pre-training. Specifically, the Knowledge-Specific strategy allows to pre-train KSCLP on project-specific bug reports corpus, by which the model can fully learn internal knowledge of bug reports, learning bug report-aware representation. As for the Contrastive Learning strategy, it performs a sequence-level pre-training for KSCLP, helping it capture the semantic information of bug reports on a global level. Upon completion of the pre-training phase, KSCLP undergoes further refinement through a Sequence-to-Sequence framework specifically tailored for bug report summarization. The efficacy of KSCLP is rigorously evaluated against five baseline models using a publicly available dataset. The empirical results demonstrate that KSCLP outperforms all baselines, achieving remarkable improvements by up to 23.73, 13.97, and 20.89 points in ROUGE-1, ROUGE-2, and ROUGE-L metrics, thereby setting new benchmarks in the field of bug report summarization.

show abstract

Section: Large Language Modelmentioning

confidence: 99%

Enhancing Bug Report Summaries Through Knowledge-Specific and Contrastive Learning Pre-Training

Shao,

Xiang

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…There are four main categories of code clones, type 1 to type 4, based on their resemblance. When there is a bug in the original code, its duplicate code is expected to contain a similar bug, causing the bugs to spread all over the software system, the same as an infection (3) . Even though fixing the original bug is feasible, the fixing process generally finds it difficult to treat all the code fragments cloned from the original code.…”

Section: Code Clone Detectionmentioning

confidence: 99%

“…To deal with issues, code clone detection-based bug detection techniques have been broadly considered and have yielded high-quality results. Therefore, code clone detection is very important as an underlying analysis technique to maintain the quality of software (1)(2)(3) .…”

Section: Code Clone Detectionmentioning

confidence: 99%

“…The token-based methods bifurcate all source code lines into a series of tokens throughout the lexical analytic step of the compiler. Next, all tokens are reconverted into token series lines (3) . The token series rows are matched to locate and inform duplicate codes.…”

Section: Code Clone Detection Techniquesmentioning

confidence: 99%

“…These identical regions form a code clone. The metric-supported clone detection methods generate separate vectors for every code piece, using metrics collected from the source code (3,4) . The areas with identical codes are discovered by comparing such vectors.…”

Section: Code Clone Detection Techniquesmentioning

confidence: 99%

See 2 more Smart Citations

Analysis of Research Trends Towards Types of Code Clone Detection Techniques

Solanki¹,

Dalal²

2023

IJST

View full text Add to dashboard Cite

Objective: The key research objectives of this study are: (1.) To compare and contrast the research trend towards the tree, token, text, metric, and graph-based code clone detection techniques; (2.) To study the distribution of metric-based code clone detection techniques on various online repositories; (3.) To make a statistical analysis of the hybrid techniques available for clone detection. The overall objective is to investigate the research trends of code clone detection approaches. Methods: Various repositories like google scholar, IEEE, and ELSEVIER Digital Libraries were systematically examined to attain the results in terms of research articles published in various places like conferences, journals, etc. followed by the inclusion and exclusion criteria. Findings: (1.) The findings related to objective 1 depicted that 50% of total clone detection techniques are tree and graph-based Code Clone Detection techniques followed by 20% of text-based and 30% of token-based code-clone detection techniques (2.) The findings related to the second objective depicted that an equal percentage of 46% of research work related to metric-based code clone detection techniques has been published in journals and conferences. (3.) The findings related to the third objective showed that 43% of hybrid code clone detection techniques are based on machine learning techniques, 24% are based on neural networks, and 18% of techniques are data mining based followed by 15% nature inspired based algorithms. Novelty: The study conducted is novel in identifying and exploring those potential code clone detection techniques that are underutilized and least explored. The result of research questions will assist researchers to draw inferences regarding usage, application, research trends, future needs, and research directions.

show abstract