FindICI: Using machine learning to detect linguistic inconsistencies between code and natural language descriptions...

Area of Science:

• Software Engineering
• Computer Science

Background:

• Linguistic anti-patterns, such as naming inconsistencies, degrade source code quality.
• Infrastructure-as-Code (IaC) scripts require high readability and maintainability for effective environment management.

Purpose of the Study:

• To develop and evaluate an automated approach for detecting linguistic anti-patterns in IaC scripts.
• Specifically, to identify inconsistencies between the implementation (body) and naming (short text names) of IaC code units.

Main Methods:

• Proposing FindICI, a novel approach utilizing word embedding and classification algorithms.
• Generating code embeddings from the abstract syntax tree (AST) of IaC code units.
• Employing machine learning techniques, including classical and deep learning models, for inconsistency detection.

Main Results:

• Evaluated FindICI on Ansible tasks extracted from open-source repositories.
• Demonstrated comparable and satisfactory performance across various word embedding models and classification algorithms.
• Successfully detected inconsistent Ansible tasks, particularly for frequently used Ansible modules.

Conclusions:

• FindICI effectively detects linguistic anti-patterns in IaC scripts.
• The proposed method enhances the maintainability and understandability of IaC code.
• Automated detection of these anti-patterns is feasible and beneficial for managing computing environments.