Semantic matching in GUI test reuse

Area of Science:

• Software Engineering
• Empirical Software Engineering
• Software Testing

Background:

• Reusing test cases across applications with shared functionalities can reduce development effort and accelerate time-to-market for reliable software.
• Existing approaches for migrating test cases across Android apps often rely on semantic matching and test generation algorithms.

Purpose of the Study:

• To define a general framework for evaluating the impact and effectiveness of different semantic matching techniques in test reuse for Android app migration.
• To conduct a comparative evaluation of various components within test migration processes.
• To propose an optimized approach by combining the most effective choices for each test migration component.

Main Methods:

• Developed a framework to systematically evaluate semantic matching and test reuse approaches for migrating test cases across Android applications.
• Conducted a comparative analysis of different semantic matching techniques, including word-level and sentence-level embeddings.
• Performed an experimental evaluation using 8,099 GUI events from 337 test configurations.

Main Results:

• Semantic matching has a prominent impact on the effectiveness of test reuse.
• Sentence-level embedding techniques significantly outperform word-level techniques for semantic matching.
• The corpus size for word embedding models showed a negligible impact on the Semantic Matching Algorithm's performance.
• Matching selected event types demonstrated superior performance compared to matching all event types.
• The overall effectiveness of test reuse is contingent upon the specific characteristics of the test suites and applications.

Conclusions:

• The choice of semantic matching significantly influences test reuse effectiveness in Android app migration.
• Sentence-level semantic matching is a more effective strategy than word-level matching.
• Future research can leverage the provided replication package to further explore and refine test reuse strategies.