Abschlussarbeit

Beschreibung:

A major challenge in the field of artificial intelligence is dealing with contradictory information. For example, when data is collected from different sources, conflicts (inconsistencies) are almost unavoidable. Classical inference fail to provide meaningful conclusions in the presence of contradictory information. To address this, novel inconsistency-tolerant, and inconsistency inspired inference relations have been proposed [1], offering a robust framework for reasoning under conflict. These relations rely on the concept of minimal hitting sets [2], which are collections of interpretations that collectively satisfy every formula in a knowledge base.

This bachelor project focuses on implementing the before mentioned inference operators using Answer Set Programming (ASP). Specifically, the project will involve designing ASP rules to encode the computation of minimal hitting sets for a given knowledge base and implementing the four novel inconsistency- tolerant inference relations introduced in the referenced study. To evaluate the effectiveness and efficiency of the implemented operators, the project will include an empirical analysis. The evaluation will assess computational performance, scalability, and the quality of results in comparison to classical and non-classical inference methods.

References

[1] Yehia Hatab, Kai Sauerwald, and Matthias Thimm. A hitting set approach to inconsistent-tolerant reasoning. In Nina Gierasimczuk and Jesse Heyninck, editors, Proceedings of the 22nd International Workshop on Nonmono- tonic Reasoning (NMR’24), November 2024.

[2] Matthias Thimm. Stream-based inconsistency measurement. International Journal of Approximate Reasoning, 68:68–87, January 2016.