Search papers, labs, and topics across Lattice.
This paper introduces First-Order Temporal Logic Tensor Networks (FOT-LTN), an innovative extension of Logic Tensor Networks that integrates linear temporal logic with fuzzy semantics to model dynamic properties and relations of objects over time. By addressing the limitations of existing neuro-symbolic AI methods that primarily focus on static knowledge, FOT-LTN enables the representation of temporal operators and quantifiers in a fully differentiable framework. The evaluation demonstrates that FOT-LTN outperforms purely neural approaches in a temporal knowledge graph completion task, highlighting its potential for advancing temporal reasoning in AI systems.
FOT-LTN outperforms traditional neural methods in temporal knowledge graph completion by effectively modeling dynamic object properties over time.
Most of the existing neuro-symbolic AI methods focus on the scenario of static knowledge where objects do not change according to a temporal dimension. Temporal neuro-symbolic works are still under explored and are mainly developed for time-interval logic or propositional linear temporal logic. There is a lack of models studying linear temporal logics with predicates that deal with objects whose properties and relations change through the time. We present First-Order Temporal Logic Tensor Networks (FOT-LTN) that is an extension of Logic Tensor Networks (LTN) that fills this gap by considering a linear-temporal dimension. In particular, FOT-LTN joins the syntax of First-Order Linear Temporal Logic with the fuzzy (and real-valued) semantics of LTN obtaining a framework that supports both temporal operators and quantifiers and is totally differentiable. A first evaluation regards a temporal knowledge graph completion task on two synthetic datasets showing better performance of FOT-LTN with respect to dedicated (purely neural) methods.
