City Research Online

Abstract

Neural-Symbolic approaches are becoming increasingly prominent due to their ability to integrate knowledge and data. In this paper, we propose the iterative use of a neurosymbolic approach and evaluate its reasoning capability. We deploy the Logic Tensor Networks neurosymbolic approach iteratively and compare its reasoning capability with purely symbolic reasoning under closed-world and open-world assumptions. Reasoning capability is evaluated on two data sets, a family relationship task and a typical ontology reasoning data set. The use of an iterative neurosymbolic approach improves reasoning from an F1 score of 0.64 to 0.97 in one case, and from 0.60 to 0.88 in the other, which is higher than what was reported previously in the literature. Our results also show that an open-world neurosymbolic approach based on differentiable fuzzy logic can excel at recall, while a logical reasoner under a closed-world assumption can achieve high precision when the domain is under-specified.

Publication Type:	Conference or Workshop Item (Paper)
Additional Information:	© 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
Publisher Keywords:	Neurosymbolic AI, Practical Reasoning, Open-World Assumption, Closed-World Assumption
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science R Medicine > RC Internal medicine > RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry
Departments:	School of Science & Technology > Computer Science

Export

Downloads