Testing software's changing features with environment-driven abstraction identification

Area of Science:

• Software Engineering
• Requirements Engineering
• Automated Software Testing

Background:

• Domain terms aid in requirements elicitation and modeling.
• Extending assistance to requirements validation is crucial for robust software development.
• Existing techniques may lack the precision needed for comprehensive testing.

Purpose of the Study:

• To present an automated approach for identifying abstractions to support requirements-based testing.
• To define novel patterns for generating testable abstractions.
• To evaluate the accuracy and effectiveness of the proposed method.

Main Methods:

• Utilized Wikipedia pages as a domain-independent corpus.
• Defined five novel patterns using part-of-speech tagging and dependency parsing.
• Framed candidate abstractions as pairs for testability.

Main Results:

• The proposed abstractions demonstrated higher accuracy compared to a state-of-the-art technique.
• The abstractions effectively revealed bugs and matched manual environmental assumptions.
• Evaluated across Electronic Health Records and Web conferencing systems.

Conclusions:

• The automated approach enhances requirements-based testing through accurate abstraction identification.
• The abstraction format improves testability and guides testing execution.
• The study proposes a new feature testing methodology focusing on environmental condition variations.