Loading...
Please wait, while we are loading the content...
Similar Documents
Improving Quality of Clone Detection with Conceptual Similarity of Source Code
| Content Provider | Semantic Scholar |
|---|---|
| Author | Pham, Hung Viet Vu, Phong Minh Nguyen, Tung Thanh |
| Copyright Year | 2017 |
| Abstract | Code clones are highly similar code fragments which are highly desirable candidates for refactoring or aspect mining. However, popular clone detection techniques sometimes report clone candidates of low quality. This paper introduces CoFi, a filtering technique to remove them. CoFi has three major steps: i) learning to represent technical terms in source code as vectors, ii) measuring the conceptual similarity of clone candidates based on those vectors, and iii) filtering out clone candidates having low conceptual similarity. Preliminary evaluation suggests that CoFi can improve detection results of popular clone detection tool. It removed 83.3% of low quality clone pairs and retains 98.6% high quality pairs from the detection result of JSync, a treebased detection tool. For DLCD, a deep learning-based detection tool, it removed 54% and retained 96.8%. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://hvpham.github.io/files/Clone-NL4SE.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
