A Neuro-Symbolic Oracle Generator
Faculty of Informatics - Academic Studies Administration
Date: 4 December 2025 / 09:30 - 12:30
USI East Campus, Room D0.10
Abstract:
Testing is essential to ensuring software quality, but defining automated test oracles remains an open challenge. This phd dissertation investigates how neuro-symbolic approaches, which combine neural learning with symbolic reasoning, can improve the automation of test oracle generation. Through a comprehensive empirical evaluation, the phd work assesses and compare neuro-symbolic, neural, and symbolic techniques for generating both axiomatic and concrete test oracles, examining their respective strengths and limitations. The results show that incorporating symbolic reasoning into learning-based techniques significantly improves the effectiveness of test-oracle generation over purely neural baselines, yielding more correct oracles while reducing the number of false positives. Overall, the research advances understanding of how neuro-symbolic integration can enhance automated software testing and software quality assurance.
Dissertation Committee:
- Prof. Mauro Pezzè, Università della Svizzera italiana, Switzerland (Research Advisor)
- Prof. Alberto Martin Lopez, Università della Svizzera italiana, Switzerland (Research co-Advisor)
- Prof. Luca Di Grazia, Università della Svizzera italiana, Switzerland (Research co-Advisor)
- Prof. Gabriele Bavota, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Carlo Alberto Furia, Università della Svizzera italiana, Switzerland (Internal Member)
- Prof. Annibale Panichella, University of Delft (External Member)
- Prof. Michail Papadakis, University of Luxembourg (External Member)