Loading...
Please wait, while we are loading the content...
Similar Documents
SpecTran : Neural Network Machine Translator of Javadoc Tags to Java Specifications
| Content Provider | Semantic Scholar |
|---|---|
| Author | Haduong, Nikita Lu, Qifan Ibtasam, Samia |
| Copyright Year | 2019 |
| Abstract | Program specifications define intended program behavior and can range in formality and completeness from natural language comments tomathematical formulations. They are important forwriting high quality, maintainable code but are difficult to write and hence often not written together with the program. Many tools that help programmers write specifications exist for specific programming languages; however, the tools often require substantial human effort to create and are language-specific. With recent advances in neural machine translation systems in the natural language processing community, we investigate whether such systems can be easily adapted to this domain of assisting programmers with creating program specifications. We test our system, which we call SpecTran, on the Javadoc informal specification scheme that is shipped with the Java programming language. Our system performs well with in-domain data, achieving up to 86.8% accuracy, but it is not robust enough to perform comparably with out-of-domain data. We provide suggestions for improving the system further and hope our work will motivate others to continue this line of research. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://courses.cs.washington.edu/courses/cse503/19sp/submissions/final-report/CSE503_Project_JavaDoc.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
