Institutional Repository of Vilnius University
Probleminio kodo aptikimas naudojant dirbtinius neuroninius tinklus /
Abstract
dc:descriptionNatural language processing models can be adapted for code analysis. This paper investigates whether natural language processing models can be used to detect problematic code – code smells. Literature relating to code smells and their detection, as well as natural language processing models is reviewed. A dataset for training, based on the QScored database, is created. GraphCodeBERT and CodeBERT natural language models, designed for source code analysis, are used to identify the following code smells: Magic Number, Complex Method, Empty Catch Block, and Multifaceted Abstraction. The results of both models are compared against alternative solutions found in the literature. The comparative analysis revealed that Magic Number and Complex Method code smells can be effectively identified by both of the analysed models; GraphCodeBERT is best for Empty Catch Block identification; neither of the models can identify the Multifaceted Abstraction code smell successfully.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Danaitis, Arnas,
- Contributors dc:contributor
-
- Uosis, Karolis
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107116361&prefLang=en_US