Institutional Repository of Vilnius University
Automatinis funkcijų pavadinimo prasmingumo tikrinimas Python programose naudojant didelius kalbos modelius /
Abstract
dc:descriptionThis thesis addresses the problem of the meaningfulness of function names in Python programs. Meaningful function names positively influence code readability and understandability. However, existing tools such as code linters and IDEs do not assess this issue and cannot evaluate whether a function name is meaningful within its context. Anyway, this problem, can potentially be solved using large language models trained to understand both natural and programming languages. In this work, the CodeBERT model was selected and further fine tuned to evaluate whether a function name corresponds to its internal context. The model was trained on a customprepared dataset and evaluated using unseen data from the CodeSearchNet dataset, functions from the PyTorch library, and through a developer survey. The results show that large language models capable of understanding both natural and programming languages can effectively tackle the challenge of evaluating function name meaning fulness. Among several models analyzed, CodeBERT was found to be the most suitable choice for this task, particularly when computational resources are limited. The fine tuned model, trained on 3,968 samples, demonstrated reasonably good performance in assessing function names. How ever, to achieve even higher accuracy, the training dataset could be expanded in both size and diversity. Further improvements to the model and its integration into CI/CD pipelines or IDEs could significantly enhance code readability, improve sostware quality, and support developer collaboration.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Minkevičius, Geraldas,
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:ELABAETD239031861&prefLang=en_US